Industry-Driven Model-Based Systems Engineering (MBSE) Workforce Competencies—An AI-Based Competency Extraction Framework

Abstract

Model-based systems engineering (MBSE) is being rapidly adopted in U.S. industries across various sectors. While practitioners and academics recognize many benefits of adopting MBSE, industries also report challenges such as limited tool expertise and a shortage of skilled personnel. Highlighting the difficulties in industry adoption of MBSE, prior research by the authors identified challenges such as tool limitations, knowledge gaps, cultural and political barriers, costs, and the level of customer understanding and acceptance of MBSE practices. Additionally, another study by the authors points out a gap between industry demands for MBSE skills in new hires and the current academic training programs. To further assess the MBSE industry’s workforce needs, this paper introduces a two-phase method for the Structured Extraction of MBSE competencies using large language models based on current workforce demands from LinkedIn job postings. Phase 1 involved extracting 1960 job descriptions from LinkedIn using the term “model-based systems engineer.” In phase 2, large language models (LLMs) employing deep transformer architectures were used to transform unstructured text into structured data. An AI agent was used as an autonomous software layer to manage every interaction between the raw dataset from Phase 1 and the LLM. Supported by the analyzed data, a competency framework is proposed that summarizes the tools, technical skills, and soft skills expected of a model-based systems engineer by the industry. The framework is designed to include core competencies shared across all MBSE roles, with specific competencies tailored for aerospace & defense, manufacturing and automotive, and software and IT sectors.

1. Introduction

Model-based systems engineering (MBSE) is a transformative approach to developing complex systems. As engineering disciplines shift toward digital transformation and integrated life cycle management, MBSE, as defined by The International Council on Systems Engineering (INCOSE), provides a structured framework to support the system development lifecycle, including systems requirements, design, analysis, verification, and validation phases, with models serving as primary artifacts [1,2]. The National Defense Industrial Association (NDIA) defines MBSE as an approach that uses models as authoritative sources of truth and includes requirements, analysis, design, implementation, and verification of a capability, system, and product throughout the acquisition life cycle [3]. Placing models at the center of system design, MBSE shifts the systems development paradigm from a document-centric to a model-centric approach.

Facilitating this paradigm is the use of modeling languages such as Systems Modeling Language (SysML), Unified Modeling Language (UML), Service-oriented architecture Modeling Language (SoAML), and Object Process Model (OPM), along with tools such as Cameo Systems Modeler, IBM Rhapsody, Simulink, and Capella. As reported across academic literature and by industry practitioners, improved traceability, early error detection, and enhanced system consistency, correctness, and completeness are the benefits of using MBSE [2,4]. Another benefit of MBSE, as noted by Hult and Stenius, is that it provides a straightforward way to document systems using established standards. This process enables validation when needed, helping to eliminate model inconsistencies and ensuring that all project stakeholders adhere to these standards. Transitioning to a digital system enhances system analysis and reduces the defects often associated with traditional, document-based methods [5]. Although several advantages have been reported regarding the use and adoption of MBSE, implementing MBSE in organizations presents several challenges. Among these challenges are a lack of user knowledge and experience in MBSE, cultural barriers, stakeholder and customer perceptions of MBSE, and limited understanding of the tools [4]. Academic research is available on MBSE methodologies, frameworks, and case studies across various sectors [6,7,8,9,10]. However, there are relatively fewer studies that concentrate specifically on workforce considerations—what skills are essential for effective MBSE adoption and how these skills are being cultivated in educational and industry settings.

The primary objective of this paper is to identify and clarify the skills, tools, and competencies required for MBSE professionals by applying natural language processing (NLP) techniques to job descriptions posted on LinkedIn. This follows a study based on previously published research by the authors, which highlighted gaps between industry requirements for MBSE skillsets in new hires and the current academic training curricula [4]. This data-driven approach enables the extraction of structured insights into current workforce needs, providing recommendations for professional development and alignment.

The rest of this paper is organized as follows: Section 2 reviews the background of MBSE and workforce development literature and explores the misalignment between academic curricula and industry needs; Section 3 details the methodology of data extraction and presents results; Section 4 discusses the implications; and Section 5 concludes with recommendations for future work.

2. Background and Literature Review

2.1. Workforce Competency Identification Using NLP and LLM

Natural Language Processing (NLP) is a powerful tool for analyzing text data online, such as job postings on the internet. These job postings often showcase the latest labor market trends and in-demand skills. By utilizing NLP, researchers can gain valuable insights into the skills, tools, qualifications, and roles that are currently in high demand across various industries. One application of NLP in workforce analytics is the extraction of keywords and their analysis in terms of frequency in job postings [11]. IT job descriptions from the Indeed platform are extracted using the Indeed REST API. Techniques such as tokenization, lowercasing, stop word removal, and stemming are employed to categorize job listings by category, title, technology, and industry sectors [11]. Pejic-Bach et al. use text mining to extract and analyze skills and competencies from Industry 4.0-related job descriptions. A combination of web scraping, term frequency analysis, word co-occurrence, and clustering techniques was used to summarize unstructured data and identify cross-disciplinary and soft skills required in the digital era [12]. Romanko and O’Mahony discuss how online job postings can provide information about job demand with a focus on creating a skills extraction process using the word2vec model, which helps identify and analyze skills mentioned in job ads. The authors employ frequency histograms of skills, scatterplots illustrating skill correlations, and graphs analyzing skill co-occurrence to gain a deeper understanding of the data [13]. Işığıçok, et al. studied 400 job postings for data scientists—200 from Turkey and 200 from the USA—utilizing text mining and created a document-term matrix to tally how often specific skills and qualifications were referenced. Subsequently, Pareto analysis was applied to pinpoint the key skills that were most prevalent, adhering to the 80/20 rule. Emphasis was also placed on skills such as data visualization, data mining, predictive analytics, NLP, and clustering techniques, all of which were deemed crucial [14]. These studies frequently employed simple methods such as regular expressions (regex) and term frequency-inverse document frequency (TF-IDF) to identify the most in-demand technical and soft skills. Although compelling for spotting patterns across vast data, these methods can struggle with ambiguity, contextual subtleties, and the diverse meanings present in job descriptions. Recent advances in large language models—specifically, transformer-based models—enable a more comprehensive analysis. Grybauskas et al. utilized the BERT Topic Model to extract insights from high-dimensional data—3828 LinkedIn postings—to identify Industry 4.0 job skills [15]. Akkasi proposed a transformer-based neural network framework to extract technical and non-technical skills from diverse job descriptions, reflecting the strength of transformer-based models in handling contextual information in natural language [16].

This paper focuses on a less-explored area within MBSE using NLP techniques to analyze LinkedIn job postings. The goal is to identify the specific skills that the workforce needs in various sectors and at different experience levels, thereby creating a workflow for extracting skills from job listings.

2.2. Current Workforce Development Needs in MBSE

As model-based systems engineering practices are widely adopted across industries, a critical issue is the growing shortage of a trained young workforce in MBSE. Reports from the industry and academia often indicate that a lack of skilled professionals is a significant obstacle to the successful application of MBSE. For example, a prior study [4] by the authors investigated the disparity between industry demands and academic preparation. It emphasizes that although MBSE offers various benefits—like enhanced system design, traceability, and lifecycle integration—its implementation faces obstacles, including limited tool familiarity, inadequate workforce training, cultural resistance, and a customer inclination towards conventional document-based approaches. Professionals from the defense, aerospace, automotive, and software industries were surveyed to determine their preferred MBSE tools, languages, and concepts for candidate selection. Findings show that tools like Cameo Systems Modeler, IBM Rhapsody, and languages like SysML and Simulink are highly regarded.

Additionally, frequently sought-after skills include knowledge of data flow diagrams, state charts, and object process methodologies [4]. Furthermore, the report on building the future systems engineering workforce, which complements the INCOSE Vision 2035, emphasizes the importance of training future model-based systems engineers, underscoring the need for a qualified workforce [17]. There is an ongoing need to explore how MBSE skills and competencies are developed in the future workforce. For instance, most engineering programs still focus on traditional systems engineering principles, offering limited emphasis on MBSE concepts. When MBSE is taught, it is often confined to elective courses or embedded within broader systems design classes. Core elements such as SysML, system architecture modeling, and tool-based design are covered inconsistently. Many universities introduce MBSE through a single course, typically at the graduate level, providing limited exposure to industry-standard tools. Additionally, educators face challenges such as restricted course time, high software costs, and the steep learning curve associated with modeling environments like SysML [4].

To effectively build a strong MBSE workforce, it is crucial to align academic curricula with industry needs through regular updates, the integration of relevant tools, and collaboration with industry stakeholders. Such efforts help equip future engineers with practical, job-ready skills while also promoting broader adoption of MBSE. Expanding upon a previous study [4], this paper examines empirical data on the skills and competencies that employers seek in model-based systems engineers. By analyzing text from job descriptions within the LinkedIn job database, we aim to inform curriculum design and professional training.

3. Methodology

A two-phase data extraction and analysis methodology is employed, leveraging both rule-based and AI-powered NLP techniques. Phase 1 involved extracting job descriptions from LinkedIn using the term “model-based systems engineer.” Phase 2 involved identifying competencies, which included extracting essential skills from each job listing through regex-based parsing, as well as gathering preferred skills, educational qualifications, and years of experience related to those skills. By leveraging the semantic extraction capabilities of LLMs and the operational efficiency of AI agents, we analyzed unstructured job descriptions to organize key information in a structured format.

3.1. Phase 1–Data Extraction

To collect real-world data on current workforce demands for MBSE, a targeted keyword search was conducted using the terms “model-based system engineer” and “ MBSE” on LinkedIn’s job search platform. LinkedIn is a professional networking platform widely used by individuals to search for jobs and, simultaneously, by companies seeking to hire a workforce. Exploring each job description in a specific field of interest manually takes a lot of time; to augment this, a LinkedIn jobs scraper library is used to extract publicly available job descriptions. This library enables automatic retrieval of job posting and their descriptions by using an API that helps set up search queries, filter job results, and extract job information. The library also enables filtering of job results by various criteria, including job type, location, and whether the job is remote.

The keywords are used to target job listings in the United States, along with applying filters that include various levels of experience, such as Associate, Entry, and Executive levels, to gather job listings. While the term “systems engineer” is used across diverse industries, it encompasses a wide range of responsibilities, spanning from gathering and analyzing requirements to integrating and managing complex systems. A traditional systems engineering role typically involves requirements management, verification and validation, systems integration, and systems architecture design activities [18]. Job descriptions of a systems engineer do not always explicitly indicate the need for proficiency in MBSE tools or practices. A model-based systems engineer primarily leverages the use of modeling tools and languages as a primary means of specifying and integrating systems across the systems engineering life cycle. Since not all systems engineering roles necessarily emphasize the need for proficiency in MBSE methodologies, the terms “MBSE” and Model Model-based Systems Engineer” are used. In total, 1960 job postings were mined, capturing essential attributes such as organization name, job title, geographic location (state and city), job level, and the full textual content of job descriptions. This dataset served as the foundation for Phase 2, which utilized NLP techniques to identify key competencies, skill demands, and role-specific trends within the MBSE domain.

Additionally, we visualized the geographic provenance of the 1960 postings at the state level (Figure 1). In this choropleth map, each state is shaded on a blue gradient proportional to its count of job listings—with darker hues indicating higher concentrations of MBSE roles—revealing that the majority of positions are clustered in California, Texas, Virginia, and a handful of other states.

3.2. Phase 2-Data Analysis

3.2.1. Initial Extraction Approach- Regex-Based Parsing

The first method used in this study involved creating a Python 3.0 script to extract structured data from job descriptions using regular expressions (regex) [19]. This approach aimed to capture key information from each job posting, including mandatory skills, preferred skills, educational qualifications, and years of experience. The process started with identifying mandatory skills. A regex pattern was applied to isolate the section of each job description between the headings “Mandatory Skills” and the next delimiter. After locating this segment, a secondary regex pattern was used to break down the extracted text into individual skill items for more detailed data collection.

A similar method was used to extract the preferred skills a candidate should have according to the employer. Here, the script used regex to find the text between the headings “Preferred Skills” and the next section delimiter. The content was then divided into individual preferred skills, following a similar approach to that for mandatory skills. For educational qualifications, regex patterns were customized to identify phrases starting with “Bachelor” and ending with “Master” (if present) or stopping at the next relevant section, like “Preferred Skills” or “Job Roles/Responsibilities.” This enabled flexible extraction, accounting for different ways qualifications are listed in various job descriptions. To capture years of experience, regex patterns were specifically designed to find and record the minimum number of years required, ensuring consistent and accurate data collection. Upon completion of these extraction steps, the process yielded a structured summary for each job description. This summary included the mandatory skills, preferred skills, educational qualifications, years of experience, and job responsibilities, providing a comprehensive and organized dataset for subsequent analysis.

Challenges and Limitations of the Regex Approach

While the regex-based extraction method provided a structured framework for parsing job descriptions, several significant challenges and limitations were observed during its implementation. One of the main obstacles was the inconsistent formatting across different job postings. Many job descriptions did not follow uniform conventions or labels, and sections such as “Mandatory Skills” or “Preferred Skills” were often omitted. Another limitation stemmed from the omission or implicit mention of key information. Important details, such as specific skills or educational qualifications, were usually implied rather than explicitly stated, making the reliable extraction of information with regular expressions alone difficult.

Additionally, the variation in language and terminology created further challenges. Job descriptions frequently used diverse phrasing, company-specific terminology, and non-standard section labels, all of which caused regular expression (regex) patterns to fail in consistently identifying relevant information. Structural ambiguity within job postings also posed a challenge. The lack of explicit delimiters between sections was especially problematic when descriptions diverged from expected structures, further complicating the extraction process. Considering these limitations, it was determined that the regex-based approach was insufficient for extracting thorough information from the wide variety of job descriptions analyzed. The significant differences in formatting and terminology across postings made this method unreliable for meeting the research goals.

3.2.2. NLP-Based Structured Extraction Using Large Language Models (LLMs)

To overcome the limitations of the regex-based extraction method, we adopted a more advanced approach using NLP techniques and LLMs [20]. This sophisticated method enabled us to analyze unstructured job descriptions and extract key information in a structured and consistent format [21,22]. The reason for choosing this approach is the combination of LLMs’ semantic extraction abilities with the efficiency of AI agents. By leveraging the deep contextual understanding of LLMs, combined with the automation and scalability of AI workflows, this method offers a robust, flexible, and scalable solution for extracting structured data from highly variable and unstructured job postings. A detailed algorithmic breakdown of this NLP- and LLM-based extraction process is provided in Section 3.2.4.

Mechanism of Entity Extraction in Large-Language Models

Large language models, such as GPT-4, utilize deep transformer architectures to convert unstructured text into structured, machine-understandable facts. This process occurs through four consecutive computational steps, combining token-level analysis with contextual reasoning.

Tokenization and Context Construction: The process starts by breaking down the raw character stream into tokens or sub-word units that can represent whole words, prefixes, or punctuation marks. For instance, the phrase “bachelor’s degree in computer science” can be split into individual tokens like Bachelor, ’s, degree, and so forth. This detailed representation allows the transformer to manage rare or morphologically complex vocabulary while keeping a fixed-size vocabulary—an essential feature for handling various job descriptions. Tokenization has been recognized as the crucial first step in natural language processing [23]. Each token is then mapped into a high-dimensional vector space and processed through multiple self-attention layers, enabling the model to grasp long-distance syntactic and semantic relationships that together create the text’s contextual meaning.

Pattern-Based Entity Recognition: During pre-training, the transformer processes billions of sentences, learning statistical patterns that connect surface forms to semantic categories. For instance, numeric expressions like “5+ years of experience” are reliably linked to specific experience requirements, whereas words such as “communication,” “leadership,” or “problem-solving” relate to soft skills. These pattern–category connections form the basis of modern named-entity recognition (NER) pipelines [24], but unlike traditional rule-based systems, LLMs perform this matching in a single forward pass through their transformer layers. Importantly, the model retrieves these associations even when job postings lack explicit section headers (e.g., “Experience” or “Required Skills”), overcoming the recall limitations of rule-based parsers and supporting robust extraction across diverse advertisement formats.

Inference of Implicit Competencies: Transformer-based LLMs often infer implied information, not just what is explicitly written. After tokenizing the input, multi-head self-attention enables each token to query all other tokens, integrating dispersed textual cues—whether within a single sentence or across an entire document—into a unified contextual representation, capturing relationships among input sequences. During pre-training, the model encounters numerous co-occurrences (e.g., “cross-functional teams” → teamwork; “taking initiative” → self-motivation), with gradient updates embedding these statistical patterns into its parameters. For inference, the decoder assigns probability mass to latent concepts that are not explicitly present in the surface strings, effectively “connecting the dots” between different parts of text [25]. A downstream prompt or task head then guides the model to verbalize these high-probability latent concepts—such as teamwork, leadership, or communication—turning implicit skills into explicit [26], structured outputs for end-user analysis.

Prompt-Conditioned Structured Decoding: Finally, an instruction-driven prompt limits the model to produce its inferences within a fixed JSON schema. The methodology’s prompt specifies necessary fields—such as soft skills, technical tools and languages, education requirements, and minimum experience—and sets “unknown” as the default for any missing data. Because both the vocabulary and output structure are predefined, the decoder produces type-safe, syntactically valid results that downstream processes can utilize without further processing. Empirical studies show that such format-specific instructions alone generate schema-compliant JSON in over 80% of cases [27]. Essentially, this prompt turns the LLM from an open-ended conversational agent into a deterministic information-extraction engine, enabling scalable and reproducible workforce analytics.

3.2.3. Role of AI Agents in Structured Extraction

In this study, the AI agent serves as an autonomous software layer that manages all interactions between the raw dataset and the LLM. Serving as the pipeline’s control panel, the agent handles prompt creation, model invocation, JSON validation, and the initiation of corrective actions when needed. This architecture converts the inherently unpredictable outputs of the LLM into dependable, analysis-ready records suitable for downstream analytics. The agent performs four critical roles:

Automation of Query–Response Cycles: The agent programmatically inserts each job description into a parameterized prompt and submits it to the LLM, capturing the generated completion and directing the response to the appropriate post-processing routine. This process eliminates manual steps from the data collection cycle, enabling seamless, end-to-end automation.

Schema Enforcement: Each model completion is analyzed and thoroughly checked against the specified JSON schema outlined in “Mechanism of Entity Extraction in Large-Language Models”. The agent confirms the presence of essential fields, validates data types, and ensures the correct use of reserved fallback tokens, such as “unknown.” Any records that do not conform are promptly corrected with instructions, ensuring that only schema-compliant objects proceed to the analytics stage.

Scalability Through Task Automation: By autonomously managing query generation, model invocation, and validation, the agent applies a consistent approach across tens, hundreds, or thousands of job descriptions. This ensures reliable extraction quality at scale, regardless of corpus size, without increasing human workload.

Error Handling and Adaptability: If a response remains invalid after re-prompting, the agent flags the record for manual review while continuing to process the rest of the queue. Both prompt templates and schema definitions are version-controlled, enabling quick adaptation when job advertisement formats change or when new competency fields need to be added.

Through these integrated functions, the agent transforms the LLM from a simple conversational tool into a reliable, repeatable extraction engine capable of supporting large-scale, dependable workforce analytics. This agent-focused approach not only simplifies data extraction but also guarantees the quality, consistency, and flexibility needed for enterprise-level information processing.

3.2.4. Algorithmic Workflow for Competency Extraction

The competency extraction pipeline, built with LangChain, incorporates a prompt template, schema validation hooks, and an AI agent into a single workflow. Figure 2 shows a visual overview, while the description below explains the five-stage process.

Input Data Preparation: The LinkedIn Jobs Scraper API retrieved 1960 U.S.-based job postings for the query “model-based systems engineer,” and “MBSE” filtered by experience level (Entry, Associate, Executive). Results were initially stored in an Excel file, and a preprocessing script removed duplicates, normalized text encodings (such as UTF-8 standardization), and generated a Python list. Each list item included a unique identifier and the raw text of a job ad, ensuring consistency for further processing.

Output Schema Definition: A canonical JSON schema was established to standardize the extracted data. The schema required five mandatory fields: Soft Skills, Technical Skills, Tools and Languages, Education, and Experience (in years). The string “Unknown” was designated as the only null token for missing information. This schema acted as the contractual interface between the LLM’s outputs and subsequent analytical processes.

LangChain Agent Configuration: A LangChain AI agent was created using LLMChain and set up to handle three primary tasks. Prompt Embedding involved inserting individual job ad text into a parameterized prompt template that specified the five target fields and the “Unknown” fallback rule. LLM Dispatch sent the completed prompt to the LLM (e.g., GPT-4) via API, and Schema Validation checked the returned JSON against a predefined schema. Notably, the prompt included an explicit formatting example, which guaranteed 100% first-pass schema compliance in this study, eliminating the need for retries.

LLM-Powered Extraction: For each job posting, the agent created a tailored prompt, which the LLM processed using its contextual understanding to identify explicit entities (e.g., “Python” → Tools & Languages) and infer hidden competencies (e.g., “collaborated with cross-functional teams” → Soft Skills: Teamwork). The LLM returned a JSON object strictly following the predefined schema, even when job descriptions lacked section headers or used non-standard terminology.

Validation and Storage: Validated JSON records were saved to a new Excel workbook, with each row representing a single job posting. Multi-valued fields (e.g., Technical Skills) were stored as comma-separated strings in their respective columns. This structured dataset enabled immediate integration with downstream analytics tools, eliminating the need for manual data cleaning.

These five stages transform unstructured LinkedIn descriptions into a clean, schema-compliant corpus—without manual annotation—enabling reproducible, large-scale workforce analytics.

3.2.5. Data Segregation Strategy

To thoroughly examine how both career seniority and industry context influence skill requirements, each job posting in the dataset was systematically categorized along two independent axes: career experience and industry sector. This two-dimensional approach enables a detailed examination of how professional advancement and industry-specific factors influence the skills employers seek, as summarized in

Table 1. Distribution of Job Descriptions by Industry Sector and Experience Level.

Industry	Experience	# of Jobs	Total %
Manufacturing & Automotive	Entry Level	76	3.88
	Associate Level	85	4.34
	Mid-Level	20	1.02
	Director Level	0	0.00
	Executive Level	1	0.05
Aerospace & Defense	Entry Level	381	19.44
	Associate Level	417	21.28
	Mid-Level	92	4.69
	Director Level	3	0.15
	Executive Level	4	0.20
Software/IT/Consulting	Entry Level	216	11.02
	Associate Level	249	12.70
	Mid-Level	57	2.91
	Director Level	1	0.05
	Executive Level	1	0.05
Other	Entry Level	144	7.35
	Associate Level	172	8.78
	Mid-Level	38	1.94
	Director Level	3	0.15
	Executive Level	0	0.00

, which displays the distribution and proportional representation of each subgroup within the 1960-posting dataset.

Experience-Based Segregation: The second column of segregation categorizes postings based on required years of experience, outlining five distinct career stages: Entry (0–2 years), Associate (2–5 years), Mid (5–10 years), Director (10–15 years), and Executive (more than 15 years). This framework illustrates the developmental path from basic technical skills to advanced strategic leadership. Notably, the dataset is most concentrated in the Entry and Associate levels, reflecting higher market demand for early-career professionals. In contrast, postings for Director and Executive roles are less common, as shown by their lower counts in Table 1.

Industry Sector Segmentation: The first column categorizes postings by industry sector, assigning each to one of four groups: Automotive & Manufacturing (including engineering, product design, and production processes), Software & IT Consulting (covering programming, cloud services, and advisory solutions), Aerospace & Defense (centered on regulatory-driven, safety-critical system modeling), and Other. The “Other” category acts as a catch-all for diverse postings that do not fit into the three main groups, ensuring complete coverage of the dataset. This sector-based tagging highlights domain-specific skill profiles while allowing flexibility for postings that span multiple or less common industries.

To improve interpretation, data are segregated by seniority to control for career-stage effects, while sector partitioning accounts for domain context. This two-way framework allows vertical comparisons of how competency develops within a single sector as roles advance, as well as horizontal comparisons of how the same career level varies across industries. Including both counts and percentages in Table 1 helps maintain awareness of cell size, thereby preventing overinterpretation of groups with small numbers.

3.2.6. Data Visualization and Analysis for Competency Identification

Data Summary

Table 2. Term clustering of relevant terms.

Industry Sector	Cluster Title	Top Terms
Aerospace & Defense	Modeling Skills	“SySML/UML”, “Cameo Systems Modeler”, “Matlab/Simulink”, “Python”, “C/C++”, “DOORS”, “DoDAF”, “UAF”, “MagicDraw”, “UPDM”, “Rhapsody”, “Java”, “Enterprise Architecture”, “CAD”, “Sateflow”
	Soft Skills	“communication”, “teamwork”, “problem-solving”, “leadership”, “collaboration”, “interpersonal”, “initiative”, “presentation”, “Analytical”, “Adaptability”, “flexibility”, “written communication”, “integrity”, “self-motivated”, “organizational”
	Tools & Technology	“Cameo Systems Modeler”, “DOORS”, “Python”, “JIRA”, “Java”, “Matlab”, “C/C++”, “Rhapsody”, “Git”, “Jenkins”, “MagicDraw”, “Confluence”,”JavaScript”, “AWS”, “SQL”
Manufacturing & Automotive	Modeling Skills	“Python”, “Matlab/Simulink”, “c/c++”, “piping & instrumentation diagrams”, “3D CAD”, “SQL”, “Simscape”, “SolidWorks”, “ASPEN”, “GD&T”, “Statistics”, “Machine Learning”, “PyTorch”, “TensorFlow”, “Scikit-learn”
	Soft Skills	“communication”, “Problem Solving”, “collaboration”, “teamwork”, “leadership”, “interpersonal”, “professionalism”, “positive attitude”, “self-starter”, “attentiveness”, “written communication”, “strong work ethic”, “adaptability”, “responsiveness”, “team player”
	Tools & Technology	“Python”, “Excel”, “PLC”, “Matlab/Simulink”, “HMI”, “SolidWorks”, “git”, “SQL”, “C/C++”, “Big Data Framework”, “scientific computing frameworks”, “CAE Tools”, “Windows OS”, “Office tools”, “Multimeter”
Software/IT/Consulting	Modeling Skills	“SysML”, “UML”, “Python”, “SQL”, “Cameo Systems Modeler”, “UAF”, “Mathlab/Simulink”, “MagicDraw”, “UPDM”, “DoDAF”, “JSON”, “BSON”, “JavaScript”, “Rhapsody”, “C/C++”
	Soft Skills	“Communication”, “Problem Solving”, “teamwork”, “Collaboration”, “leadership”, “analytical”, “project management”, “Self Starter”, “Presentation”, “Critical Thinking”, “Mentorship”, “Interpersonal”, “Organizational”, “Integrity”, “Creativity”
	Tools & Technology	“Python”, “JIRA”, “Git”, “Kubernetes”, “DOORS”, “AWS”, “Java”, “MagicDraw”, “Docker”, “Cameo”, “MATLAB”, “C/C++”, “Rhapsody”, “Linux”, “Spark”
Other	Modeling Skills	“SySML”, “MATLAB/Simulink”, “Python”, “UML”, “SQL”, “AutoCad”, “C/C++”, “SolidWorks”, “CAD”, “Java”, “Cameo Enterprise Architecture”, “UPDM2/UAF”, “Octave”, “COMSOL”, “GT-Suite”
	Soft Skills	“Communications”, “Teamwork”, “Problem Solving”, “Leadership”, “Collaboration”, “Analytical”, “Interpersonal”, “Written Communication”, “Self Motivated”, “Verbal Communication”, “Project Management”, “Mentorship”, “Attention to detail”, “Adaptability”, “Organizational”
	Tools & Technology	“Python”,”DOORs”, “Cameo Systems Modeler”, “MATLAB”, “Java”, “Git”, “AWS”, “JIRA”, “C/C++”, “Azure”, “Excel”, “MagicDraw”, “Jenkins”, “Kubernetes”, “Oscilloscopes”

provides a summary of the top-ranked terms for each industry sector, grouped into three main competency categories: Modelling Languages, Soft Skills, and Tools & Technologies. These terms were identified using TF–IDF scoring, which highlights the most critical and sector-specific skills, tools, and attributes in the analysis. By organizing the most frequent items into a single matrix, the table converts qualitative patterns observed in the dataset into a precise, numerical reference. This tabular format enables quick cross-sector comparison and offers a clear lexical snapshot of workforce demand, maintaining the specificity of each sector while allowing systematic analysis across domains. A notable point from Table 2 is the repeated occurrence of certain terms across various industries and clusters. For instance, programming languages like “Python” and modeling tools such as “Cameo Systems Modeler” appear in both Aerospace & Defense and Automotive & Manufacturing, as well as in multiple clusters (e.g., Modelling Languages and Tools & Technologies). This underscores their centrality in contemporary MBSE practices. Likewise, soft skills like “teamwork,” “communication,” and “leadership” are prominent across all sectors, emphasizing their widespread importance in the workforce. This repetition underscores both the cross-cutting relevance of specific technical and interpersonal skills and the overlapping demands of modern engineering and technology roles. This finding highlights a crucial need for a balanced combination of technical and interpersonal training in MBSE.

The tabular summary thus serves as both a complement and a refinement to earlier qualitative visualizations, offering an exact, research-ready account of the skills landscape within and across industry sectors. By capturing both unique and recurring terms, the table supports nuanced workforce analytics and informs targeted upskilling or recruitment strategies.

Python and Cameo Systems Modeler: Both Python and Cameo Systems Modeler consistently appear across various sectors, particularly in Aerospace & Defense and Automotive & Manufacturing, serving as both modeling languages and essential tools. This overlap underscores their versatile functions as tools for simulation and as general-purpose programming platforms. The dual presence of these competencies indicates their value across various stages of the MBSE process, including requirements analysis, model development, testing, and automation of tasks such as scripting, data analysis, and tool integration.

Simulink and MATLAB: The presence of Simulink and MATLAB across all sectors indicates that they are critical modeling environments, pivotal for simulation and analytical processes. Their capability to handle complex, dynamic systems modeling and simulation is relevant in sectors demanding rigorous validation and verification processes for safety-critical systems.

Computer-Aided Design (CAD): CAD is widely used in the Automotive and Manufacturing Industries, as well as in aerospace and defense, where it is crucial for product lifecycle management, system visualization, and detailed engineering design. The connection between CAD and MBSE could be closely related to CAD’s vital role in design-focused industries that require precise graphical models as core elements of the MBSE approach.

Cloud Computing and Object-Oriented Programming: Cloud computing and object-oriented programming (OOP) languages are prominently used in the Software and IT consulting sector. This highlights the industry’s emphasis on flexible, scalable solutions and software-focused system design.

The frequent appearance of Python and Cameo Systems Modeler highlights a crucial skill set that is widely valuable for MBSE roles, strongly recommending that curricula and training programs heavily incorporate these competencies to prepare adaptable MBSE professionals. The overlap between Simulink and MATLAB across various industry sectors highlights the need for targeted training in simulation and analytical software within engineering education, emphasizing the importance of practical, hands-on experience with these specific tools. The shared use of CAD tools underscores the importance of integrating comprehensive design and visualization skills into MBSE training, mirroring real-world industry demands. The notable presence of Cloud Computing and OOP in software-focused sectors suggests that MBSE training should include modules or projects centered on modern software development practices, cloud-based architecture, and scalable system designs to stay aligned with current technological trends.

Analysis of Technical Skills and Tools

To quantitatively compare the emphasis on modeling languages and engineering tools across different industry sectors, we created a 4 (industry sectors) × L (skills/tools) matrix based on the proportional frequencies of each term in the corpus. Each column was then converted to a Z-score, where positive values indicate above-average emphasis and negative values indicate below-average emphasis for that skill or tool within its distribution. This normalization enables meaningful comparisons between sectors by removing scale effects related to sector size or overall skill frequency. Figure 3 shows the heat map. Warm color (Dark blue) highlights sectors that emphasize a tool or modeling language more than the average across all industries. Cool colors (such as cool blue) indicate sectors that place less emphasis than average on a particular skill or tool. For example, if the cell for “Python” in the Software sector is a deep red and shows a high positive Z-score, this indicates that Python is mentioned much more frequently in Software & IT Consulting job postings than in other sectors. Conversely, a blue cell for “CAD” in Software indicates it is relatively less emphasized there.

In the software and IT-related sectors, it is observed that Python, cloud computing, object-oriented programming, and the use of SysML are highly emphasized, reflecting the digital and service-oriented nature of the industry. The automotive and manufacturing sectors place a greater emphasis on utilizing Python, MATLAB, Simulink, and CAD, highlighting their focus on mechanical and system design. In contrast, the aerospace and defense industry’s preference for Cameo Systems Modeler and MATLAB demonstrates significant prominence, underscoring the industry’s reliance on rigorous model-based engineering practices.

The technical skills heatmap highlights sector-specific emphasis on tools and languages. Notably, Simulink and CAD tools are prominent in sectors dedicated to tangible product design and manufacturing, whereas the Software and IT sectors emphasize programming languages like Python and cloud services. The heatmap not only clarifies current demand but also helps inform the future systems engineering workforce, academia, and the expected technical skills needed by industry for MBSE professionals, aiding in aligning MBSE educational content with specific industry needs. Figure 4 represents tools and technical skills emphasized across various MBSE experience levels. This heatmap highlights MATLAB, Atlassian/JIRA, Python, DOORS, and C as key tools across different experience levels. MATLAB and DOORS are particularly prominent at the Entry and Associate levels, underscoring their fundamental roles in modeling and requirements management. Python and C are increasingly significant, particularly at the Mid-Level, showcasing their importance in simulation and software integration. While tools like Cameo Systems Modeler/MagicDraw are also relevant in MBSE practices, they currently have a moderate emphasis.

Analysis of Soft Skills

A 4 (industry sectors) × L (Soft Skills) matrix was created for soft skill terms, with each column standardized to Z-scores. Figure 5 illustrates these standardized values, allowing for direct comparison of behavioral skill priorities across industries. Warm cells (reds) highlight sectors where a soft skill is particularly emphasized compared to others, and cool cells (blues) show where a soft skill is less prioritized. For instance, if the skill “Communication” appears as a consistently warm cell across all sectors, it indicates that this competency is universally valued.

Communication and Teamwork appear as consistently warm cells, indicating their widespread importance across all sectors. Leadership, Problem-Solving, and interpersonal skills show the highest intensities in the Aerospace & Defense sector. The Automotive industry shows elevated Z-scores for organizational and performance-focused skills, while the Software sector leans toward Innovation and Adaptability.

The soft skills heatmap reveals a consistent focus across industries on interpersonal skills, particularly in Communication and Teamwork. Notably, the Aerospace & Defense sectors emphasize Leadership and Problem-Solving, reflecting the complexity of their projects and high-stakes environments. This heatmap provides actionable insights for educators and industry trainers to prioritize these universally critical and sector-specific skills in their curricula and training programs.

Figure 6 represents the soft skills emphasized across various MBSE experience levels. Communication, Interpersonal Skills, and Initiative are the most consistently valued soft skills across Entry-Level, Associate-Level, and Mid-Level MBSE roles. Communication remains a key focus at all experience levels, highlighting its essential importance. Likewise, Interpersonal Skills and taking Initiatives are also persistently highly regarded.

4. Discussion

This paper presents empirical insights into the essential competencies for MBSE professionals, derived through advanced NLP techniques applied to MBSE industry job descriptions. The data extraction and Analysis conducted highlights a blend of technical proficiency, particularly in tools like Simulink, Cameo Systems Modeler, and Python, along with soft skills such as Communication and Teamwork. The innovative aspect of the study lies in the application of sophisticated NLP methodologies to systematically identify real-world MBSE competencies from extensive job market data.

Drawing insights from the analysis presented, key competencies required by industries in a model-based systems engineer include proficiency in system modeling using SysML and UML, programming and simulation skills in tools such as MATLAB, Python, Simulink, and Cameo Systems Modeler, as well as familiarity with concepts of Cloud Computing and Object-Oriented Programming principles. Additionally, the analysis emphasizes the importance of soft skills, including Communication, Teamwork, Leadership, Problem-Solving, Interpersonal Skills, Organizational Skills, Performance Orientation, Innovation, and Adaptability, which are essential across various industry sectors for a model-based systems engineer.

Figure 7 illustrates a competency framework that summarizes the tools, technical skills, and soft skills expected of a model-based systems engineer by industry, allowing industry professionals, educators, and workforce planners to quickly identify and adapt to changing skill requirements. It provides clear, sector-specific profiles based on thorough analytical processes, acting as a reference to guide the workforce on essential technical and soft skills that align with current industry standards.

In addition to the core competencies and industry-specific competencies identified, a thorough analysis of the data extracted helped identify additional supporting competencies such as familiarity with Object-Oriented Programming, proficiency in Cloud Services, ability to utilize Data analysis and visualization tools, requirements management skills, familiarity with verification and validation methodologies, and systems integration and testing skills needed for a model-based systems engineer. The frequent appearance of skills such as Python, MATLAB, Simulink, and Cameo Systems Modeler across various sectors highlights their essential and versatile roles within the broader MBSE field. This insight confirms the strength of the competency framework and underscores the widespread importance of specific skills and tools in the MBSE workforce.

It is recognized that MBSE adoption varies considerably across industries, with some sectors, such as Aerospace & Defense, having well-established MBSE roles, while others, like Automotive, are still in the early stages of adoption [4,28], which naturally affects how explicitly MBSE roles are outlined in job descriptions. In the automotive industry, it is standard practice to have a central authoritative database that houses 3D CAD data to assess manufacturing, servicing, and performance requirements for modeling product systems. An MBSE environment for product development would function similarly to product lifecycle management, serving as a single source of truth for attributes, dimensions, and other related specific representations of a system of interest in CAD. In addition to CAD representations, the system models would enable the behavioral aspect of the system to evaluate its feasibility [29]. Consequently, automotive job postings may reflect hybrid roles that blend MBSE responsibilities with traditional engineering design tasks, such as CAD usage. This observed overlap aligns with findings from existing literature [28,29], suggesting that MBSE adoption often involves broader engineering tasks to support gradual transition and integration of MBSE practices into established processes. Therefore, it is reasonable to assume that competency descriptions from the Aerospace & Defense and Software sectors more clearly indicate mature MBSE roles, whereas Automotive sector postings tend to reflect transitional and hybrid roles.

Limitation: Dissecting Table 2 and Figure 2 and Figure 3 reveals overlaps in competencies, particularly in modeling languages and tools across different sectors, which can be attributed to the underlying NLP and computational methods employed. LLMs, such as GPT-4, recognize semantic similarities and contextual relationships between competencies across different industry sectors’ job descriptions. Competencies that appear within similar contextual environments or job descriptions are highlighted and grouped due to their semantic proximity and frequent co-occurrence. Additionally, the use of a structured extraction method, where the LLM converts unstructured text into structured JSON outputs, also contributes to overlapping competencies. This extraction method explicitly delineates categories such as “Technical Skills” and “Tools & Technologies.” Skills and tools that are commonly used interchangeably or within similar contexts may appear prominently across multiple sectors, emphasizing genuine cross-sector applicability rather than methodological artifacts.

5. Conclusions and Future Work

Aligning education and professional development closely with identified industry needs is essential for addressing skill gaps, supporting workforce readiness, and fostering the growth and effectiveness of the MBSE workforce in dynamic industrial environments. A total of 1960 job postings were mined, capturing the full textual content of job descriptions from the LinkedIn Platform. To analyze the extracted data, an advanced approach using NLP techniques and LLMs was employed. This method allowed for the analysis of unstructured job descriptions and the extraction of key information into a structured, consistent format. By leveraging the deep contextual understanding of LLMs along with the automation and scalability of AI workflows, this approach provided a robust, flexible, and scalable solution for extracting structured data from highly variable and unstructured job postings. A detailed analysis of job descriptions highlights the importance of incorporating specific technical skills into MBSE training programs. A set of core technical competencies, observed to be common to all MBSE job roles (based on the extracted data), along with industry sector-specific competencies, has been identified. Proficiency in Python and Cameo Systems Modeler is highly sought after due to their widespread use in the industry, suggesting that academic curricula should focus on these tools.

Additionally, skills in Simulink and MATLAB are essential for hands-on simulation and analytical tasks, underscoring their importance for systems simulation and analysis. The demand for detailed design and visualization abilities through CAD tools reflects the use of MBSE for mechanical and detailed system design. By highlighting the key competencies, the illustrated competency model can proactively address skill gaps, improve workforce preparedness, and support the growth and ongoing relevance of MBSE practices across various industries. This competency model can be used to integrate specific MBSE training into engineering curricula at undergraduate and graduate levels, utilizing the identified core competencies such as modeling proficiency in SysML, essential programming, and key skills. It incorporates MBSE modeling tools, such as Cameo Systems Modeler, MATLAB, and Simulink, which align with real-world industry demands. Furthermore, academic institutions can create specialized elective courses and certifications tailored to the specific needs of industries like Aerospace & Defense, Automotive & Manufacturing, and Software & IT Consulting. Future work will include enhancing the competency framework by incorporating data from similar job posting databases, such as Indeed, Glassdoor, and CareerBuilder, as well as from career centers of professional organizations like INCOSE and IISE.