AI Engineer (LLM & Voice AI)
Summary
Hands-on AI Engineer building production LLM integrations, AI agents, and real-time voice agents (STT→LLM→TTS) for a logistics-industry platform — owning latency budgets, eval harnesses, and prompt versioning. Requires 4+ years software engineering, strong Python, and fluent English/Lithuanian. Vilnius-based, hybrid.
About the Role
On behalf of our client, DATAHEAD is looking for a hands-on AI Engineer to build AI-powered solutions for the logistics industry.
You will work with operational data, including GPS, vehicle telemetry, routes, driver activity, and transport orders, to develop intelligent solutions that help transport companies reduce costs, improve fleet efficiency, and automate daily operations.
The role focuses on LLM integrations, AI agents, real-time voice systems, and production-ready AI applications.
What you will do:
• Design and ship LLM integrations into the platform: tool/function calling, structured outputs, RAG where it earns its place.
• Build and tune real-time voice agents (STT → LLM → TTS): streaming, barge-in, turn detection, endpointing, interruption handling.
• Own the latency budget end to end - measure time-to-first-token, time-to-first-audio and turn latency per pipeline stage, and cut it.
• Build evaluation harnesses: golden datasets, regression suites for prompts, LLM-as-judge where justified, and voice-specific metrics (turn-taking errors, false interruptions, WER, task completion).
• Write and version prompts like code: tests, diffs, rollbacks.
• Make model/vendor decisions with data (Claude, OpenAI, open-weight; Deepgram, ElevenLabs, etc.)
Must haves:
• 4+ years software engineering, of which 1+ years shipping LLM-based features to production users.
• Hands-on prompt engineering for production: structured outputs, tool calling, multi-turn agents, failure-mode handling.
• Built at least one real-time or low-latency LLM system (voice agent, streaming chat, live assistant) and can explain where every 100 ms went.
• Designed and ran LLM evaluations -offline datasets and online metrics - and used them to make release decisions.
• Solid Python: idiomatic, readable, maintainable code - clear naming, small focused functions, type hints, tests, and code reviewers can follow without a walkthrough.
• Solid fundamentals: async/streaming I/O, WebSockets/WebRTC, queues, observability.
• Fluent in English and Lithuanian
Vilnius-based office and hybrid work possibilities.
Gross salary range is 5000- 6500 EUR/month.
This search is being delivered exclusively by DATAHEAD.
Apply directly or message me confidentially to discuss.
Skills
Sign in to read and join the discussion.