Applied AI Engineer
I build LLM systems
that hold up in production
Nine years of software engineering, the last three in applied AI. I design and run agentic systems end to end — harnesses and tool orchestration, MCP integrations, RAG, evals and observability: the parts that turn a demo into a product.
Prefer email? [email protected]
Expertise
What I work on
Production LLM systems
Design, build, and operate LLM applications that real users rely on every day — with reliability, latency, and cost as part of the spec, not an afterthought.
Agentic systems
Multi-step agents with tool use: orchestration, harness code, guardrails, and failure handling that keep an agent on task.
MCP integrations
Model Context Protocol servers and integrations that expose products, data, and services to AI agents.
RAG
Retrieval-augmented generation over real corpora: indexing, retrieval quality, grounding, and honest measurement of answer quality.
Evals
Reproducible evaluation pipelines that gate every prompt and model change. Quality as a number, not an opinion.
LLM observability
Tracing, cost and latency monitoring with Langfuse: knowing what the system actually does in production instead of guessing.
Experience
Where this comes from
NOW
Applied AI engineer at a major travel company
I build and operate a production LLM assistant: agentic harness and tool orchestration, MCP integrations, retrieval over domain data, an evals pipeline that gates changes, and Langfuse-based observability. I own the path from prototype to production.
BEFORE
Nine years of software engineering
Backend and product engineering, mostly in Python — the production habits that LLM systems need came first, the models second.
About
Product thinking, engineering accountability
I take problems from “could an LLM do this?” to a system running in production — and answer for the result.
If a problem is better solved without an LLM, I say so early. Honest evals beat impressive demos.
I work hands-on, end to end — the person who scopes the problem is the one who ships and runs the system.
Python · LangGraph · MCP · RAG · evals · Langfuse
Contact
Get in touch
Telegram is the fastest way to reach me; email works too.
I’m also open to consulting engagements around production LLM systems.