Jeremy W.

I'm a Senior Python and AI engineer with 10 years of experience. I add LLM and RAG features to products that already have users, plus the backends, evaluation and monitoring that keep them reliable. WHAT I BUILD MOST AI into products that already exist. At a European telephony company I shipped LLM call summarization into a live platform, used prompt folding, meta-prompting and fine-tuning to hold output quality after moving to a cheaper LLM model to maximize client profits. I built the evaluation layer with DeepEval, scoring summaries on faithfulness, hallucination, relevancy and contextual precision, so quality stayed measurable while the prompts kept changing. RAG and conversational systems. Retrieval pipelines over a medical device knowledge base, queried by voice by technicians working hands-free in surgical suites, on constrained hardware and under HIPAA. Backends that carry real load. Event-driven pipelines on AWS using Lambda, ECS, SQS with dead-letter reprocessing and EventBridge, sustaining delivery across millions of records, with self-healing retries and a custom observability suite that caught defects the platform's own monitoring missed. I am also comfortable telling you what not to build. Scope questions are cheaper to answer in week one than in month three. STACK Python, Django, Go, FastAPI, Flask, PostgreSQL, Celery, Redis, AWS, GCP, Docker, REST APIs, Kubernetes, React and Next.js. LLM integration, RAG, vector search, prompt engineering, evaluation and observability. ON UPWORK SINCE 2018 I tend to keep long-running relationships with clients and at 4000+ hours across 15 contracts, I have a 100% Job Success Score. If you have an idea for a product but need input on the architecture and implementation, or a product that needs AI in it, or an AI feature that works in testing and breaks in front of users, that is the work I do most and I would be happy to talk it through with you at any time!