Tarek Masryo tarekmasryo

AI engineering focus areas

I build, evaluate, and operate production ML and generative AI systems across model serving, RAG, AI agents, evaluation, and LLMOps.

Featured Work

Project	Engineering focus
Evagix · PyPI	Local-first repository evidence validation and governance CLI with deterministic checks, traceable findings, JSON/SARIF output, and validation behavior refined across 100+ repositories
RAG QA Command Center	CI-tested RAG evaluation workspace spanning 3,824 offline QA evaluation runs and 93,375 retrieval events, with retrieval diagnostics, hallucination-risk analysis, trace inspection, configuration comparison, and regression analysis
LLMOps Telemetry Command Center	Operator-facing LLMOps workspace analyzing 9,000 offline interactions across 1,595 sessions, with diagnostics for reliability, latency, cost, routing behavior, drift signals, and operational review
Fraud Risk Ops Platform	End-to-end fraud risk operations platform combining calibrated ML models, policy-driven decisions, FastAPI inference, analyst review, audit trails, background processing, and Prometheus/Grafana observability; achieved F1 0.882 for Random Forest and 0.876 for XGBoost

Area	Engineering scope
Production ML & Generative AI Systems	Model serving, inference workflows, APIs, background processing, containerized delivery, testing, deployment, monitoring, and operational integration
RAG & Retrieval Systems	Ingestion, embeddings, vector and hybrid retrieval, reranking, retrieval evaluation, source attribution, grounded outputs, structured responses, and quality controls
AI Agents & Agentic Workflows	Tool use, code execution, stateful workflows, validation, controlled execution, approvals, retries, failure recovery, executable checks, and behavior analysis
Evaluation & Benchmarking	Multilingual benchmark design, model/LLM/agent/retrieval evaluation, executable checks, regression testing, trace analysis, model comparison, and failure diagnosis
Observability, Reliability & LLMOps	Telemetry, tracing, monitoring, latency/cost analysis, routing behavior, drift signals, quality monitoring, failure visibility, and operational triage
Applied ML & Decision Systems	Tabular ML, NLP, computer vision, calibration, threshold optimization, error analysis, explainability, risk scoring, and operator-facing decision support

Area	Technologies
Programming & Backend	Python · SQL · FastAPI · Pydantic v2 · SQLAlchemy · Alembic · PostgreSQL · Redis · REST APIs · Linux · Bash · Git
Machine Learning & Data	PyTorch · TensorFlow/Keras · scikit-learn · XGBoost · LightGBM · Pandas · NumPy · Polars · DuckDB
Generative AI, RAG & Retrieval	Hugging Face Transformers · Sentence Transformers · Embeddings · Vector Search · Hybrid Retrieval · Reranking · pgvector · FAISS · Context Engineering · LangChain · OpenAI API · Ollama · vLLM
Agents, Evaluation & Benchmarking	LangGraph · MCP · Tool Calling · Code Execution · LLM Evaluation · Agent Evaluation · Retrieval Evaluation · RAGAS · DeepEval · Benchmarking · Failure Analysis
MLOps, Testing & Infrastructure	Docker · Docker Compose · RQ · MLflow · GitHub Actions · CI/CD · pytest · Ruff · mypy · Schema Validation · Model Serving & Inference · Caddy/Nginx
Observability & Reliability	OpenTelemetry · Prometheus · Grafana · Structured Logging · Tracing · Model Monitoring · Latency/Cost Monitoring
Frontend & AI Interfaces	TypeScript · JavaScript · React · Tailwind CSS · Streamlit · Gradio

Applied ML, analytics, data products, and supporting projects

Open to AI engineering roles, remote contracts, and selected technical collaborations across:

Production ML & generative AI systems — model serving, APIs, deployment, evaluation, observability, reliability, and system integration
RAG, agents & AI evaluation — retrieval quality, grounded outputs, agentic workflows, tool execution, benchmark design, trace analysis, and failure diagnosis
Applied ML & decision systems — model evaluation, calibration, threshold policies, monitoring, analytics, and operator-facing workflows

Best contact: LinkedIn