10/25/2026 updated
AT
100 % available
AI engineer: production LLM and voice agents, evals, full-stack (Python/TypeScript)
Amsterdam, Netherlands
Only remote
MEng, Electrical & Computer Engineering, National Technical University of AthensAbout me
AI and software engineer in Amsterdam, freelancing since 05/2025. Most of my work is the part after the demo: getting AI systems into daily use. Former CTO of a voice-agent startup. Remote, EU time zones, one month's notice.
TensorFlowTerraform
AI agents & evaluation: production voice and LLM agents (ElevenLabs, Pipecat, Claude Agent SDK, MCP, LangChain/LangGraph); eval harnesses with simulated users and LLM judges; LangWatch.
Real-time voice: streaming speech-to-text and TTS pipelines (Deepgram, Azure Neural TTS, Cartesia); translation with domain glossaries (DeepL).
Backend & full-stack: Python (FastAPI, Django), TypeScript (React, Next.js, Prisma), Postgres.
Cloud & delivery: Terraform on Azure, Azure Container Apps, GitHub Actions, Docker; coding-agent pipelines (Claude Code) with CI gates and human review.
Earlier: time-series forecasting and anomaly detection (TensorFlow), telecom network monitoring (C/C++), cloud-to-edge research (Java, Kubernetes, TOSCA).
Experience
Professional experience0 - 5 years
Freelancer since2025
Languages
EnglishFluentGreekNative speaker
Project history
Built and run a risk-assessment platform, a live captioning tool for multilingual site crews, and document automation, working directly with the client's managers and site foremen. Stack: FastAPI, React, TypeScript, Terraform, Azure.
CALT is an event-discovery product grown to over 5,000 users, plus Cue, which matches events with sponsors.
Owned the production phone agent that troubleshoots EV chargers for charge-point operators: 20K chargers under support within six months, 70% of calls resolved with no human. Built its evaluation harness (LLM-simulated callers, LLM judge, mocked backends) and rebuilt the agent on Pipecat. Stack: ElevenLabs, Pipecat, Python, MCP.