Journey

Michael Sweatt

Senior Systems Engineer

Pittsburgh, PA

Requirements engineering and system safety leader with experience on certified, safety-critical platforms across rail, automotive, and locomotive domains. Currently lead requirements engineering for the rail industry's first cloud-based safety-critical train dispatch platform, authoring and governing a 500+ requirement subsystem specification with lifecycle traceability in Polarion. Prior decade at GM spanning ISO 26262 hazard analysis, FTA/SEFA, DFMEA, and fail-operational design across the BEV propulsion portfolio. Published empirical researcher on AI evaluation reliability in engineering workflows. Built and piloted an AI-augmented requirements-to-design pipeline (Polarion → version-controlled artifacts → traced wireframes → working prototype) now being rolled out across teams through the engineering AI Workgroup. Proven leading cross-functional teams of 30–50 engineers.

01 · Experience

Siemens Mobility

1 role
April 2025 Present
April 2025 Present

Senior Systems Engineer

Requirements engineering lead for a cloud-based, safety-critical train dispatch platform, moving safety-critical rail dispatch into a cloud environment.

  • Author and own the Subsystem Requirements Specification (SSRS): 500+ requirements decomposed from the system-level SRS and defining how the dispatch platform manages trains, consumed downstream by the subsystem design specification, UI/UX specification, and customer-facing ICD
  • Govern bidirectional traceability in Polarion from operational hazards through software-level safety requirements to verification: 400+ requirements refined from the system-level SRS, 300+ linked to test cases to date on the active program
  • Decompose system hazards into software safety requirements for a distributed cloud architecture, working daily with software developers, the safety assurance team, testers, and the customer
  • Validate operational requirements directly with customer dispatch stakeholders through interactive prototype demonstrations; reduced requirement validation cycles (weeks to days)
  • Architected an AI-augmented requirements-to-design pipeline: read-only MCP extraction of Polarion specifications into version-controlled, machine-readable artifacts, driving requirement-traced UI wireframes and a working clickable prototype generated from the wireframes alone. Compressed requirements to wireframes from weeks to days on the pilot subsystem
  • Surfaced specification defects (contradictory requirements, unspecified states, happy-path-only coverage) before implementation by designing every screen against the extracted requirement set
  • Author and maintain the committed agent skills with structural guardrails (read-only by construction, boundary tripwires, content-hash drift detection)

General Motors

4 roles
August 2015 April 2025
June 2021 April 2025

Lead Systems Technical Specialist

Provided technical leadership to 40+ propulsion system architects and engineers across GM's BEV portfolio (Cadillac Lyriq, Optiq, Silverado EV, Hummer EV).

  • Conducted hazard analyses, Fault Tree Analyses (FTA), and Single Element Failure Analyses (SEFA) feeding SFMEAs and system safety assessments for high-voltage propulsion under ISO 26262
  • Authored and decomposed safety requirements for fail-operational and degraded-mode strategies: single-motor limp-home for dual-motor configurations, half-pack operation under battery cell faults, and active discharge (HV bus below 60 V within 300 ms of key-off)
  • Led cross-functional NACS charging diagnostic investigation (20+ SMEs, 8-week sprint) to root cause of fleet-wide charging failures traced to a stale legacy requirement, avoiding a potential recall across four nameplates; block and sequence diagrams adopted as standard DFMEA artifacts
  • Reduced critical system failures in BEV propulsion to the 48 IPTV target through structured DFMEA methodology; mentored systems engineers in requirements development and DFMEA analysis
September 2020 June 2021

Advanced Battery Algorithm Engineer

Led team implementing battery diagnostics and prognostics enabling preemptive recalls of vehicles with potential fire hazards.

  • Developed and validated battery life metric algorithms for dual-pack architectures in MATLAB/Simulink
  • Accelerated software development 25% via SIL methodology
June 2017 September 2020

BEV Lead Systems Engineer

Authored Propulsion Subsystem Specifications and interface control documents for the EV portfolio.

  • Led 15+ member integration teams and 50+ member product development forums
  • Drove design decisions through system DFMEAs and program reviews
August 2015 June 2017

Hybrid Systems Engineer

Built foundational requirements-at-scale infrastructure for GM propulsion programs.

  • Spearheaded migration of requirement specifications from Excel to IBM DOORS, designing a requirements management process adopted by subsequent propulsion programs
  • Led European BEV feasibility study coordinating GM Europe and Streetscooter third-party vehicle integration

GE Transportation

2 roles
October 2009 August 2015
September 2011 August 2015

Auxiliary Power Subsystem Team Lead

Directed auxiliary power distribution and battery charging systems for locomotives.

  • Integral contributor to the GE Evolution Series Tier 4 launch (65–85% emissions reduction)
  • Developed charging control strategies in C and MATLAB/Simulink
October 2009 September 2011

Edison Engineering Development Program

Competitive two-year rotational program: four engineering assignments plus graduate coursework applied toward M.S. at Georgia Tech.

  • Completed 4 rotations across Hardware-in-the-Loop Simulation, Engineering Quality, Diesel Engine Control, and Trip Optimizer

02 · Education

Master of Science in Electrical Engineering

December 2013
Georgia Institute of Technology

Bachelor of Science in Electrical Engineering

May 2009
University of Maryland College Park

03 · Certifications

Project Management Professional (PMP)

Project Management Institute

DFSS Black Belt

General Motors

Deep Learning Specialization

Stanford University

Machine Learning Specialization

Stanford University

04 · Skills

Requirements Engineering

Requirements authorship & decomposition at scale (system → subsystem → software)Safety requirements traceabilityPolarion (daily, MCP-integrated)IBM DOORS/DNGEnterprise ArchitectUML/SysMLRequirements process design

Functional Safety

ISO 26262 (Parts 3–6: HARA-adjacent hazard analysis, FSR/TSR decomposition, software safety requirements)BS EN 62267GERT8000FTASEFA/SFMEADFMEAFail-operational & degraded-mode design

AI-Augmented Engineering

Claude CodeMCP integrations (Polarion, Figma)Agent skill authoring & guardrail designRAG architecturesLLM evaluation design (pre-registered evals, judge calibration)Token-economics governancePrompt engineering

Languages & Tools

PythonMATLAB/SimulinkC/C++ (familiar)GitDockerLinuxSIL/HIL testingCI/CD

05 · Selected Projects & Research

CraftRole

craftrole.com

Shipped an AI-powered career-discovery platform with a four-person team. Developed the repo-as-context operating model here first (authoritative CLAUDE.md, committed spec-generation and UX-audit skills, doc-lint, spec-first PRs).

Next.jsSupabaseAnthropic APIClaude CodeVercel

AI-Assisted Requirements Decomposition System

Multi-agent system automating requirements decomposition workflows: 34% quality improvement over a manual baseline with 100% output consistency across 110 tests. Presented to Siemens engineering leadership as a proof of concept for AI-augmented requirements engineering.

AI agentsRequirements engineeringEvaluation

The Meter and the Judge

Read the study

Pre-registered empirical study (~4,000 measured API calls) on cost governance and LLM-as-judge reliability in AI-augmented engineering workflows. Task cost is stable and bandable, but judge–human agreement on quality was 54.9%, with 83% of disagreement traced to underspecified quality criteria.

LLM evaluationPre-registered studyCost governance

Requirements Engineering Evals

Ongoing pre-registered evaluation framework for AI detection of requirement defects in safety-critical corpora, with a frozen corpus and human-only seed authorship protocols.

LLM evaluationSafety-criticalPre-registration