I build AI systems for production: APIs, dashboards, and monitoring to improve operations, reduce risk, and scale reliably.
Production AI systems: APIs, dashboards, and monitoring, built for evaluation, reliability, and observability.
Hi, I’m Tarek Masryo,
an AI/ML Engineer building decision-ready, evaluation-first systems teams can rely on.
I ship measurable impact through APIs, dashboards, and monitoring.
What I deliver
Decision-ready AI systems with deployable artifacts, evaluation, and monitoring.
Production deliverables built for reliability, evaluation, and observability.
Prediction and risk-scoring systems shipped as deployable APIs with calibrated thresholds, operating policies, and success criteria.
Grounded RAG and tool-calling workflows with retrieval evaluation, schema-validated outputs, and pragmatic safeguards (timeouts, retries, fallbacks).
Decision-ready dashboards that translate model outputs into actions, KPIs, and human review workflows.
Robust data ingestion with schema contracts and validation to prevent silent failures and regressions.
Incident-ready telemetry covering quality signals, drift indicators, latency/cost, and alerting to catch issues early.
Core stack first. Expand to view the full stack.
Systems + Scripting
Wrangling + Notebooks
Modeling + Training
Retrieval + Tool Use
Storage + Retrieval
Serving + Validation
Interactive Delivery
Packaging + Release
Testing and tooling
Frequently asked questions about delivery, evaluation, and support.
Production ML and GenAI systems for real-world decision-making—risk scoring, forecasting, NLP/CV, and grounded RAG/agent workflows—chosen to fit practical constraints.
I pick the simplest approach that meets the target: classical ML, deep learning, or LLM/RAG only when it adds measurable value under cost, latency, and reliability constraints.
Yes. I handle end-to-end delivery: data pipelines, modeling, evaluation, backend APIs, deployment, and handoff—built to run reliably, not as a demo.
Leakage-safe evaluation, calibration and thresholding tied to cost/risk, plus error slicing and automated regression tests to catch failure modes before and after release.
Reliability and reproducibility: data contracts, schema validation, typed interfaces, versioned artifacts, and monitoring signals that surface drift and performance decay.
Clean, maintainable code with clear structure, tests where it matters, and documentation that captures assumptions, decisions, and how to run or extend the system.
Scope alignment → a written plan (risks, milestones, success criteria) → iterative delivery in focused sprints with validation checkpoints → release and stabilization.
Yes. Optional post-delivery support focused on stability, performance, monitoring, and safe iteration as requirements and data evolve.
Let’s connect on LinkedIn. Open to production AI roles and high-impact freelance engagements.