Applied AI / LLM Engineer
About Me
I am an AI / LLM Engineer focused on building LLM features and shipping them to production. My work centers on RAG, semantic retrieval, cost-aware model routing, output control, and self-hosted or local inference.
I build systems that are designed to be reliable end-to-end. In my own production projects, I have implemented intent classification, prompt caching, persistent memory, and regression-tested routing between models such as Haiku and Sonnet.
I have six years of production background in support, QA, and monitoring, which gives me a strong operational mindset. That experience helps me think carefully about reliability, observability, and how AI systems behave in real-world environments.
I have also worked on client-facing AI integration projects, where I owned the LLM layer across multiple products. This included retrieval and embedding tuning, prompt design, model selection, API integration, and monitoring with tools like Langfuse and Grafana.
My recent work includes Telegram AI agents, job-search automation, and creative-writing routers, all running in production on infrastructure I manage myself. I am comfortable working across Python, FastAPI, PostgreSQL, React, Docker, and Linux, and I enjoy building practical systems that solve real product problems.
I am looking for an AI / LLM Engineer role where I can own LLM behavior and reliability end-to-end in a fast-moving product team. I am especially interested in RAG, routing, output control, evaluation pipelines, and the surrounding systems that make AI products dependable.
Skills
PythonSQLReactDockerLinuxGitPostgreSQLMonitoringGrafanaTechnical Writing
Education
Advertising & Public Relations
Foreign Philology (English, German & Literature), Teaching Qualification
Experience
Remote product IT company. Technical owner of the LLM layer across client projects: RAG system design, retrieval and embedding tuning, prompts, output control, model selection and routing, LLM API integration, and monitoring. Designed a support and analytics platform with semantic search across 4 data streams; directed embedding-model migration and improved retrieval quality to 0.82 from 0.59. Designed an AI sales and negotiation coach and wrote its development spec. Designed the rebuild of an assistant-platform bot into a single AI agent. Led decomposition, backlog, roadmap, sprints, specs, estimates, and acceptance across projects.
Integrated LLMs into existing products and built Telegram bots from scratch with AI integration for small-business and founder clients under NDA. Completed intensive independent study of LLMs, RAG, agent design, and inference-cost optimization, including a local-LLM sandbox on networked Raspberry Pi.
Designed the prompt layer and knowledge base for an LLM support assistant that autonomously resolves about 73% of L1 tickets. Built device telemetry across a 10k+ device fleet, Grafana dashboards, and Telegram alerting. Performed QA, smoke, regression, and A/B testing; handled edge-device flashing over SSH; mentored a mid-level engineer.
Worked in L2 support and monitoring on a high-traffic real-time platform with about 15-minute response SLA. Used Graylog, Grafana, Telegram alerts, and REST API diagnostics. Also handled L1-L3 support, project management, CI from the support side, omnichannel B2B/B2C support, manual QA, moderation, and built Chrome extensions and knowledge bases.