ML Architect

Erol Akarsu

Richmond, VA *******@*****.*** linkedin.com/in/erol-akarsu github.com/eakarsu

U.S. Citizen Eligible for a U.S. security clearance (not currently cleared) Discovery Engineer — AI/ML Evaluation, Validation & Edge Deployment SUMMARY

Principal Software Engineer and ML Architect (Ph.D., Computer Science, Syracuse) with 25+ years of production software experience (12+ writing production-grade Python systems and pipelines) and a specialty in exactly this mission: proving AI works before it ships. At Lowe's, leads transformer ML and generative-AI systems with evaluation gates, human review, and model routing. Personally built and inspected 939 standalone and 204 grouped applications, including a governed AI- workload deployment platform that releases models only after measured memory, latency, throughput, power, and error checks; an agent governance control tower with tool-call auditing and deployment approval; and model-operations tooling aligned to NIST AI RMF. Builds reusable evaluation pipelines, regression harnesses, and monitoring dashboards; defines metrics with systematic historic capture; and documents results for readiness-style reviews. EVALUATION OWNERSHIP — EVIDENCE

End-to-end evaluation infrastructure: evaluation gates, typed grounded-answer contracts, reviewer approval, and append-only evidence records across Company Brain (156 files) and the AI Agent Governance Control Tower (125 files: agent registration, permission policies, tool-call auditing, human approval, rollback, incident review).

Edge/resource-constrained validation: ChipProfiler / inference-chips-for-agent-workflows (153 files, 8 test files) — immutable release authoritative hardware profile measured placement evaluation independent approval/override, with bounded canaries and telemetry-based promotion/rollback.

Metrics discipline: model routing with catalog/evaluation versions, budget reservations, quality floors, and circuit breakers (AI Model Cost Orchestrator, 128 files); monitoring dashboards and KPI instrumentation in operational platforms; systematic per-run records in ai_runs-style audit tables.

Agentic AI assessment: governed agent systems evaluated through tasking, decision, and audit trails — permission-aware answers, prompt-injection quarantine, schema-validated output, and human-in-the-loop release decisions; AI Governance/ModelOps platform organized around EU AI Act, NIST AI RMF, and ISO 42001 (138 files).

Simulation & scenario environments (adjacent): AI3DSpatial 3D/spatial platform (126 files); drone operations and sensor- fusion applications (counter-swarm-defense, 161 files) exercising scenario-based behavior over multi-source telemetry; digital-twin-adjacent test tooling in the portfolio. Classified simulation environments are not claimed.

Test harnesses & CI: automated unit/integration/route suites (state machine, signed-webhook, AI-analyzer tests — 19/19 passing in the dropshipping platform work), Jenkins CI/CD, Kubernetes deployment, and regression discipline across 1,100+ built applications.

Troubleshooting & program ownership: 25 years debugging distributed production systems (5M-record pipeline 360 40 minutes); roadmaps and readiness-style reviews with senior leaders; mentoring and cross-team coordination. TECHNICAL TOOLKIT

Python (primary, 12+ years production), Go and C++ for test automation, harnesses, and tooling; TypeScript/JavaScript, Java, Scala, SQL; FastAPI, Node, React/Next.js, Spark/Beam, Solr/Elasticsearch, PostgreSQL, ChromaDB.

MLOps & delivery: model/evaluation versioning, canary and rollback automation, Kubernetes, Docker, Terraform, Jenkins; GCP and AWS.

ML: transformers, retrieval/RAG and grounded generation, deep-learning and RL coursework (advanced algorithms, unsupervised/recommenders/RL).

PROFESSIONAL EXPERIENCE

Lead Machine Learning Architect and Principal Software Engineer Lowe's — Oct 2020–Present (ML lead since Jan 2024)

Own evaluation and release criteria for transformer image-selection and generative assistance; near-real-time data pipelines; mentoring and cross-functional roadmaps. Senior Software Architect Wipro — 2018–2020 · Master Software Engineer Capital One — 2016–2017

Production data/search platforms on GCP (Terraform/Packer/Ansible/Jenkins); automated API testing from specifications, graph-modeled dependencies, regression suites. Independent Consultant — 2003–2017 · Ferguson & Cisco — 1999–2016

Government/commercial systems end to end; secure service routing; distributed search (Nutch/Hadoop/Solr) with explicit ranking evaluation.

EDUCATION

Ph.D., Computer Science (IDE for Grid Computing), Syracuse University, 1994–1999; M.S., Computer Science, Syracuse University; M.S., Computer Engineering, Istanbul Technical University; B.S., Computer Engineering, Ege University.

Contact this candidate