ML/AI engineer · full-stack delivery

Turning advanced AI ideas into products that teams trust.

I build models, retrieval systems, and backend workflows that move from experimental notebooks into reliable production software. My focus is practical AI: clear interfaces, measurable quality, and systems that keep working when real-world constraints show up.

View work Start a build GitHub

RAG + Agents Eval Loops Prod APIs Reliability

Current build loop

Agentic RAG systems

Designing retrieval and orchestration patterns that stay grounded under production load.

Input layer

PDFs, structured records, and event streams normalized for retrieval.

Reasoning layer

Tool-using agents with guardrails, fallback policies, and audit logs.

Quality layer

Coverage checks, human review loops, and regression prompts for drift.

pipeline.log


              ingest -> chunk -> embed -> retrieve -> synthesize

              verify citations -> rank confidence -> deliver summary

              monitor latency, cost, and answer quality by session

Signature style

Engineering rigor with product clarity.

Grounded by design

Retrieval-first architecture with explicit traceability and citation-aware outputs.

Built to operate

Latency, monitoring, and failure modes treated as first-class product requirements.

Readable to humans

Clear UX and message structure so technical depth remains approachable.

Workflow

How delivery moves from idea to production.

Step 01

Scope the outcome

Define user jobs, quality bar, and constraints before model choices.

Step 02

Build the core loop

Implement ingestion, retrieval, orchestration, and structured response flow.

Step 03

Instrument quality

Add evaluation checks, error analysis, and human feedback controls.

Step 04

Scale with confidence

Ship observability and reliability paths that support long-term iteration.

Featured case study

RAG Summarizer

A retrieval-augmented system that condenses long research documents into readable, structured summaries with transparent source grounding.

RAG Retrieval Python LLM Ops

Read case study View repo

Evaluation checklist

Citation consistency across sections
Coverage of key technical claims
Readable structure for fast scanning

System map · runtime path

Ingest Retrieve Summarize Validate

More work

Additional highlights

Medical education

Meridian

A medical report reader for educational use that parses PDF medical reports, integrates Apple Watch exports, captures chronic illness histories, and helps patients understand recent events in context.

Launch workspace Contact for demo

Research topic

ANN vs. CNN: Feature Engineering vs. Deep Learning for Automated Coin Grading

A comparative study on the Saint-Gaudens Double Eagle gold coin, evaluating feature engineering + ANN baselines versus end-to-end CNNs for automated grading, with emphasis on practical deployment tradeoffs.

Read on arXiv View all projects

Disclaimer: Meridian is for educational purposes only and does not provide medical advice. Always consult a qualified healthcare professional.

Contact

If you are building serious AI, I can help ship it.

I work across the stack to turn ML ideas into reliable products: retrieval + agents, backend APIs, evaluation loops, and production readiness.

What I can do

Full-stack ML delivery

RAG + agent systems with evals, guardrails, and tool use.
Backend integration with APIs, pipelines, and clean data contracts.
Production scaling across latency, cost, monitoring, and reliability.

Helpful context

User profile and the job to be done.
Data sources, shape, and expected volume.
Timeline, constraints, and success criteria.

Direct

Fastest ways to reach me

Email is best. I usually reply with a concrete next-step plan in 24-48 hours.

Email me Resume

GitHub

Typical response: 24-48 hours.