Machine Learning Engineer · Nashville, TN
I’m Sam McDevitt. For 4+ years I’ve built and shipped machine learning in regulated, high-stakes environments — computer vision and sensor fusion for DHS and DoD, medical imaging models in healthcare, and secure agentic LLM systems — on a foundation of safety-critical embedded engineering.
Selected Work
Deployed for government field trials, clinical review pipelines, and live operations — each one owned from data pipeline to production.
U.S. Customs & Border Protection
Designed and deployed a production computer-vision system — detection plus model ensembling over X-ray, camera, and CT modalities — that scored 98.9% on the government’s technical challenge and became the direct basis for a $25M, five-year production contract. Fused X-ray imagery, structured text, and vehicle records in a cloud-native AWS pipeline for live field trials.
Medical Imaging · AI Metrics
Led modernization of the company’s core AI codebase — framework selection, migration strategy, and measurable gains in segmentation Dice and regression R² over the legacy baseline. Improved a production classifier to cut required radiologist corrections by 20%, and architected AWS infrastructure for distributed hyperparameter tuning in a regulated healthcare environment.
Real-time Sensor Fusion
Engineered a real-time system fusing Intel depth cameras with radiation detectors to attribute and track radioactive sources to their carriers. Built explainability into the deployment pipeline — Grad-CAM, RISE, hierarchical occlusion — so non-technical field operators could trust and adopt the models.
U.S. Air Force · SBIR
Led prototype development for two U.S. Air Force SBIR contracts — $2.5M combined — and both demos secured follow-on production deployments. Owned the full model lifecycle: proof-of-concept, vehicle-region segmentation paired with anomaly classification, through production with versioning, monitoring, and CI/CD.
Experience
A deliberate arc: safety-critical embedded systems first, production machine learning built on top of that discipline.
Jun 2025 — Present
Leading the modernization of the core medical-imaging AI codebase and the team’s ML practice. Production preprocessing pipelines across multi-modal medical datasets, distributed tuning on AWS, and translating clinical and regulatory requirements into model acceptance criteria with physicians and business stakeholders.
May 2023 — May 2025
Computer vision and sensor fusion for DHS and DoD. Shipped the CBP detection system behind a $25M award, led two winning Air Force SBIR prototypes, optimized models for real-time edge inference, and mentored junior engineers on MLOps — versioning, monitoring, CI/CD.
Aug 2022 — May 2023
Developed and verified mission-critical embedded algorithms for PAC-3 (Patriot) missile seeker systems in Ada, VHDL, and MATLAB — requirements engineering, verification testing, and documentation in a safety-critical defense environment.
Aug 2021 — Aug 2022
VHDL development and test for satellite communications systems; contributed to a hypersonic missile software program; built a Kalman-filter-based control system in Python for a seeker testbed.
Aug 2020 — Aug 2021
CNN segmentation models in TensorFlow/Keras for autonomous-driving perception; multi-modal sensor fusion — camera, LiDAR, depth — for ROS-based autonomous vehicle demonstrations.
Clearances U.S. Secret (previously held — Boeing/Dynetics) · DHS (previously held — CBP contract) · eligible for reinstatement
Featured Projects
Personal systems I design, deploy, and operate end to end — one live in production right now, both open on GitHub.
Agentic, access-controlled, federated retrieval-augmented generation over live PubMed — air-gap-capable, exposed over MCP. Hand-built, no LLM framework.
NCAA Division I win-probability and analytics platform — twelve stacked models behind a self-hosted dashboard that runs live all season.
Technical Skills
Education
Publication
McDevitt S, et al. “Wearables for Biomechanical Performance Optimization and Risk Assessment in Industrial and Sports Applications.” Bioengineering (Basel), 2022.
Contact