AI Architect – Voice AI

This is a fully remote position, open to applicants in Poland.

📋 Description

• Take charge of the technical architecture and implementation of the voice copilot, transitioning from a validated proof of concept to full production.

• Achieve targets related to latency, accuracy, concurrency, and cost.

• Maintain production quality expectations and prevent any unnoticed scope expansions.

• Continuously share knowledge with the client team and Neurons Lab engineers.

• Manage the entire pipeline, including streaming speech-to-text, LLM field extraction, Chrome extension delivery, and AWS infrastructure.

• Aim to reduce P95 latency from about 6 seconds to 2 seconds and address post-processing lag corner cases.

• Conduct A/B tests between the Claude Haiku and GPT Luna models, employing golden-set evaluations for phonetic name and email accuracy.

• Supervise Langfuse traces, accuracy dashboards, live-call per-call cost assessments, and optimization strategies.

• Strengthen the system to support 5–10+ concurrent calls with strict user data isolation, monitoring, alerting, and safe rollback procedures.

• Deliver comprehensive epics from start to finish, including the Amazon SES email briefing service, while ensuring a demo fallback is available.

• Facilitate technical discussions with the client and showcase measurable system performance.

• Connect feedback to the statement of work (SOW) and direct roadmap items to upcoming phases.

• Ensure that client-facing materials successfully pass the ADM review process.

• Lead the AI Engineering team and pod by delegating tasks, reviewing outcomes, and removing obstacles.

• Quickly assimilate the outgoing architect's handover to become self-sufficient.

• Conduct knowledge-transfer sessions to eliminate any single points of failure.

• Provide estimates and architectural options for the production SOW upon request.

⛳️ Requirements

• Hands-on experience with real-time voice pipelines, including streaming STT, turn handling, and low-latency LLM inference.

• Experience in LLM engineering, such as prompt engineering, structured extraction, guardrails, and model A/B testing.

• Familiarity with observability and evaluations, including tools like Langfuse, golden datasets, and dashboards for latency, accuracy, and cost.

• AWS expertise, including Bedrock, serverless design patterns, SES, token economics, and per-call cost engineering.

• Proficient in Python programming.

• Adequate knowledge of TypeScript and Chrome extensions to oversee the integration process.

• Excellent spoken and written English to communicate effectively with demanding US executives.

• Understanding of contact-center and agent-assist patterns and metrics, such as handle time, cost per call, and concurrency.

• Experience with production LLM operations, including load testing, data isolation, and incident response.

• Proven track record in deploying Voice AI in production; must have launched at least one real-time voice or speech product for actual users.

• A minimum of 6 years of hands-on AI/ML engineering experience, with strong recent involvement in LLM production.

• Demonstrated record of latency and reliability, including measurable P95 reductions and fixes for concurrency issues on live systems.

• Senior consulting/client-facing experience, capable of managing detailed UAT scrutiny and expectations.

• Preferred: experience in empathy-sensitive domains like healthcare, veterinary care, or insurance; familiarity with PE-sponsored rollouts; knowledge of Chrome extension delivery; telephony/streaming technologies like Amazon Connect, Twilio, or LiveKit; practical experience with Langfuse; US client experience with Eastern-time overlap.

• Required English proficiency level: Intermediate, Upper Intermediate, or Advanced options available.

• Availability for part-time remote work.

• Willingness to collaborate as a B2B contractor.

🏝️ Benefits

• Multi-month contract with a high likelihood of extension.

• Minimum of 0.5 FTE, increasing towards 1.0 FTE as production scales.

• Structured handover process and knowledge transfer from the outgoing architect.

• Chance to contribute to a production-stage AI program for a significant US private equity-backed client.