Profile
Senior AI engineer focused on production LLM systems and applied AI. Builds agentic workflows, model integrations, evaluation approaches, and application backends end-to-end, from technical discovery and architecture through implementation and verification. Experienced working directly with clients; classical ML and MLOps provide a supporting foundation.
Experience
- Own architecture and implementation of agentic AI workflows spanning LLM orchestration, tool integrations, APIs, data, and application services.
- Lead technical discovery, feasibility and estimation, and technical interviews; turn ambiguous client requirements into buildable solutions.
- Design LLM evaluations across task quality, reliability, tool use, latency, cost, and context handling, informing client and internal product architecture.
- Built functional agentic systems for property management and lead acquisition using Claude/Anthropic, OpenAI, Gemini, PydanticAI, FastAPI, React/Next.js, and PostgreSQL.
- Built energy-cost forecasting for 3,000 stores, producing monthly and yearly forecasts by store, region, and country with blended tree-based, statistical, and recurrent models.
- Deliver end-to-end applied AI engagements, from problem framing and data pipelines through model evaluation, deployment, and client handoff; focused on AI and LLM systems since 2023.
- Built and deployed a Gemini-based vehicle-detailing quoting workflow.
- Selected engagements: Google enterprise work through a subcontractor; NVIDIA DLSS algorithm testing and error isolation; Linux MLOps environments for Red Hat and Canonical; algorithm fine-tuning for Teralytics.
Selected Projects
Open source
Batteries-included Pi package for an always-on workflow: clarification, governed planning, bounded implementation, verification, durable .ged/ memory, repo-map awareness, skill discovery, and optional subagent staffing.
TypeScript · Node.js · Pi
Open source
Desktop and web coding workspace with a project-manager agent, isolated Git worktrees, tracked task pipeline, human approval gates, durable state, recovery, and Codex, Claude, and OpenCode providers. Personally designed and implemented the Orchestrator end-to-end.
TypeScript · React · Electron · SQLite
Open source
User-guided Chrome/Edge browser assistant for web research and automation, with page extraction, form interaction, tab navigation, and export to markdown, sheets, PDF, or HTML. Forked from badlogic/sitegeist.
TypeScript · Chrome/Edge extension · AI tools
Education
BSc Computer Science, First-Class Honours
2021 – 2025
Lancaster University Leipzig
Thesis: compared Q-learning, DQN, Dueling DQN, and Monte Carlo Tree Search in a 4×4×4 Tic-Tac-Toe environment.
Skills
AI systems
Production LLM systems · agentic workflows · tool calling · structured outputs · prompt engineering · LLM evaluation · model experimentation · API integration
Languages
Python · TypeScript/JavaScript · SQL · Rust · Go · Java · C · Lua
Frameworks / infra
FastAPI · PydanticAI · React · Next.js · Electron · PyTorch · PostgreSQL · SQLite · Docker · Linux · Git · CI/CD · GCP · Terraform · MLOps
Models / services
Claude/Anthropic · OpenAI · Gemini · ElevenLabs · Twilio
Certifications: Microsoft Certified: Azure AI Fundamentals (AI-900) · Anthropic API Certification
Languages: German (Native) · English (C2) · Romanian (Fluent)