Hi, I am
I build LLM-powered products, RAG systems and autonomous AI agents. My job is to make models do the right thing with the right context.
As an AI Engineer and AI Context Manager, I work on systems that turn large language models into real products. My focus is making LLMs produce reliable, measurable and sustainable output — not just impressive demos. What started with prompt engineering today covers context engineering, retrieval architectures and multi-step agent systems.
RAG pipelines, vector-based information retrieval and tool-using autonomous agents are at the core of my daily work. I believe the outcome is shaped not only by what a model says, but by which context it is given, in which order and within which budget — which is why I treat context design as its own engineering discipline.
I have built production integrations with the OpenAI and Anthropic ecosystems; evaluation (eval) processes, cost optimization and guardrail design are a natural part of my projects. The goal is simple: turn AI from a marketing line into the hardest-working member of the team.
Capabilities grouped by area, with the projects behind each one. The related-projects link takes you straight to that work.
I design not just what the model says, but what it says it based on.
System prompt design, few-shot strategies and structured output.
Context budgeting, memory design and dynamic context assembly.
Streaming and function calling with the OpenAI and Anthropic APIs.
Test sets and regression checks that measure output quality.
I move the right information into the context, at the right time and in the right amount.
Production pipelines with chunking, embeddings, hybrid search and reranking.
Indexing and search infrastructure with Qdrant, pgvector and Pinecone.
Document parsing, cleaning and metadata enrichment.
Flows that plan, use tools and know when to stop.
Autonomous agent architectures that carry out multi-step tasks.
Least-privilege, single-purpose tool design and error handling.
Human approval, output validation and cost limits.
From demo to production: services, interfaces and monitoring.
AI backends, data pipelines and LLM services.
Streaming chat interfaces and AI-powered UIs.
Server-side development for content and automation flows.
Discovery
Use case, data sources and success criteria are made explicit.
Measurement
An eval set and baseline metrics are defined so improvement becomes measurable.
Development
Retrieval, prompting and agent flows are built in small steps.
Monitoring
Cost, latency and quality are tracked in production.
Production-focused AI systems: the problem, the approach and the outcome. Click any card to read the full case study.
4 projects
An assistant that answers questions in natural language over corporate PDFs and knowledge bases, always citing its sources.
An assistant that answers questions in natural language over corporate PDFs and knowledge bases, always citing its sources.
An orchestration layer that automates email triage, calendar planning and reporting with tool-using agents.
From brief to publication: a production line combining research, drafting, editorial review and SEO checks in one flow.
A memory service based on summarization and vector recall that keeps long-running chats consistent.
Have an AI project, a consulting question or an idea to collaborate on? Reach me through any channel below — I usually reply the same day.