Lead Voice AI Architect for Production Co-Pilot

Neurons Lab; Poland; Hybrid; PLN 300,000 - 420,000; Full time

Job summary

Neurons Lab is building a production-grade voice copilot for a leading US veterinary care network, taking a PoC into live production. You will own the full architecture—streaming STT, LLM field extraction, Chrome-extension delivery, and AWS infrastructure.

You will drive latency reductions, ensure strict data isolation, and run production tests with real metrics. You will mentor the AI Engineer and lead knowledge transfer, with a structured handover from the outgoing architect to ramp quickly.

Qualifications

• Real-time voice pipelines and streaming STT experience.
• LLM engineering including prompt design and extraction.
• Observability, dashboards and cost/latency tracking.
• Hands-on AWS Bedrock/serverless and SES knowledge.
• Full-stack pragmatism with Python and some TS/Chrome extension.
• Strong written and spoken English for US executives.

Responsibilities

• Own the full pipeline from streaming STT to LLM extraction and extension delivery.
• Drive latency improvements and reduce P95 latency.
• Run model A/B tests with golden-set evaluation for accuracy.
• Own evaluation, dashboards and per-call cost analysis and optimization.
• Ensure production readiness with multi-concurrent calls and data isolation.
• Deliver end-to-end epics with demo fallbacks and production polish.

Skills

Real-time voice pipelines

LLM engineering

Observability and evals

AWS Bedrock

Full-stack pragmatism

English communication

Tools

Langfuse

Chrome extension

AWS SES

Job description

Neurons Lab is building a production-grade voice copilot for a leading US veterinary care network, taking a PoC into live production. You will own the full architecture—streaming STT, LLM field extraction, Chrome-extension delivery, and AWS infrastructure.

You will drive latency reductions, ensure strict data isolation, and run production tests with real metrics. You will mentor the AI Engineer and lead knowledge transfer, with a structured handover from the outgoing architect to ramp quickly.