Quality engineering at GM, Tesla, and Shield AI → self-taught ML → building production LLM systems on cutting-edge hardware. Every project ships with CI/CD, type hints, structured logging, and real deployment — not notebooks.
Currently open to ML/AI Engineer roles in Dallas–Fort Worth.
1,880+ tests · 34 W&B runs · 5,925 training steps · fp8 Blackwell inference
| Project | What It Does | Technical Highlights |
|---|---|---|
| bible-ai-assistant | Fine-tuned Bible Q&A with voice | ORPO-aligned Qwen3.5-4B, hybrid RAG (BM25 + dense + cross-encoder), constitutional guardrails · 183 tests, 34 W&B runs |
| godspeed-coding-agent | Security-first open-source coding agent | 4-tier deny-first permissions, hash-chained audit trail, speculative tool dispatch, 200+ LLM providers via LiteLLM · 854 tests |
| sovereign-edge | Privacy-first multi-agent AI on Jetson Orin | 5 LangGraph agents, ONNX DistilBERT intent router (<10ms), LiteLLM gateway (4 providers + Ollama), SOPS+Age secrets · 393 tests |
| gpu-server-test-suite | GPU fleet diagnostics modeled on NVIDIA DCGM | 16 diagnostic modules, Prometheus + Grafana, fault injection, JUnit XML for CI · 188 tests |
| manna-trading | Multi-agent AI crypto trading | 4-agent pipeline (TA → Chief → Risk → Execution), DeepSeek R1 reasoning, Kelly Criterion, Monte Carlo · 35 tests |
| llm-wiki | AI knowledge base — Karpathy's LLM Wiki pattern | LangGraph ingest/query pipelines, instructor + Pydantic structured output, BM25 search, Groq→Gemini→Ollama fallback · 117 tests |
| ml-lab | ML experiment control plane | Experiment lifecycle, model registry (append-only JSONL), cloud training launcher (RunPod/Lambda), W&B integration · 28 tests |
| manufacturing-quality-analytics | Semiconductor NCR analysis pipeline | SQL + Python ETL, supplier performance scoring, defect Pareto, yield trend analysis, PostgreSQL · 24 tests |
LLM / Inference — vLLM · TensorRT-LLM · Ollama · Unsloth · HuggingFace Transformers · ExLlamaV2
Training — PyTorch · TRL (SFT, ORPO, GRPO) · QLoRA · PEFT · Weights & Biases · Blackwell fp8
Agents — LangGraph · MCP · LiteLLM · instructor + Pydantic · LlamaIndex
Infrastructure — FastAPI · Docker · PostgreSQL · Prometheus · CUDA 13.2 · Jetson Orin Nano
Dallas–Fort Worth, TX · omnipotence-eth.github.io