I turn business problems into AI systems that run in production.

My work sits at the point where a business problem becomes a running AI system. I have spent the last few years building retrieval-augmented assistants, multi-agent workflows, LLM applications and machine-learning pipelines for enterprise and product teams — and shipping them as containerised services that people rely on every day.

I started in applied computer vision, building shipping-label detection and OCR pipelines for logistics, then moved into machine learning platforms and agentic automation at Bacancy. There I built a no-code ML framework that let non-technical teams run regression, forecasting and anomaly detection in a single click, and an AI sales agent that automated the outbound funnel from lead discovery to booked meetings.

Today, as an AI/ML Software Engineer II at ProductSquads, I design RAG systems over US regulatory and safety standards so product teams can query compliance requirements conversationally instead of reading hundreds of pages. That means owning the hard parts: ingestion and chunking strategy, embedding and retrieval quality, prompt design, evaluation and integration with legacy systems.

Alongside employment I take on independent work with international clients — from RLHF response-quality evaluation for LLM training to QA, documentation and product demonstrations for a US-based AI product company. It keeps me close to how AI products are judged, tested and explained to real users.

I care about production readiness over demos: observability, cost, latency, failure modes and honest evaluation. I experiment constantly, but I ship deliberately.