AI Architect (Voice AI)
Pubblicato il 30-10-2026 - Neurons Lab in Roma
ph3About The Project (description, Duration, Stage) /h3pThe client is the largest US network of in-home veterinary hospice and end-of-life care. A major US private-equity sponsor drives the AI program and plans more projects across its portfolio. /ppWe built a real-time voice copilot for their Veterinary Care Coordinators (VCCs). The copilot listens to live calls with pet families. It extracts appointment and clinical fields while the call runs. It fills the client's scheduling system through a Chrome extension. A second workstream, the bVet Visit Copilot /b, sends each vet an AI pre-visit briefing by email (Amazon SES). /ppNext is the bproduction phase /b. /ppbStage: /b production SOW in executive alignment; start expected September 2026. /ppbDuration: /b multi-month, with strong extension probability. 0.5 FTE minimum; ramp toward 1.0 FTE as production scales. /ppbWhy the role is open: /b the current architect moves to another strategic build. He stays at b0.15–0.2 FTE for supervision and knowledge transfer /b during ramp-up, so the new architect gets a structured handover. /ph3Objective /h3ulliOwn the technical architecture and delivery of the voice copilot from validated PoC to production /liliHit the bar this client tests against: latency, accuracy, concurrency, and cost /liliKeep expectations aligned: production polish is in scope now; protect the team from silent scope creep /liliTransfer knowledge continuously to the client's team and Neurons Lab engineers /li /ulh3Areas of Responsibility /h3h3Technical architecture hands-on implementation /h3ulliOwn the full pipeline: streaming speech-to-text, LLM field extraction, Chrome-extension delivery, and AWS infrastructure /liliDrive latency work: cut P95 from ~6s toward ~2s; remove post-processing corner cases (occasional ~1min lag on one field type)
/liliRun model A/B tests (current pair: Claude Haiku vs GPT Luna) with golden-set evaluation for phonetic name and email accuracy /liliOwn evaluation and cost: Langfuse traces, accuracy dashboards, real per-call cost from live calls, and an optimization plan /liliHarden for production: 5–10+ concurrent calls, strict data isolation between users, monitoring, alerting, and safe rollback /liliShip epics end to end (example: the SES email briefing service); always keep a demo fallback so a live session never fails /li /ulh3Working with client stakeholders /h3ulliFront technical discussions with a meticulous client; VCCs test edge cases and expect production quality /liliPresent concrete system behavior, with numbers — this account rewards evidence, not slides /liliHold the scope line: tie every feedback item to the SOW; route roadmap items (learning loop, persistent memory) to future phases /liliKeep internal discussions internal; all client-facing materials pass ADM review before sending /li /ulh3Team knowledge /h3ulliLead the AI Engineer and the pod: set tasks, review output, unblock fast /liliAbsorb the handover from the outgoing architect (0.15–0.2 FTE supervision window)
and become independent fast /liliRun knowledge-transfer sessions; the project must have no single point of failure /liliSupport the production SOW with estimates and architecture options when the account team asks /li /ulh3Skills /h3ulliReal-time voice pipelines: streaming STT, turn handling, low-latency LLM inference — hands-on /liliLLM engineering: prompt engineering, structured extraction, guardrails, model A/B evaluation /liliObservability and evals: Langfuse or similar; golden datasets; latency, accuracy, and cost dashboards /liliAWS: Bedrock, serverless patterns, SES; token economics and per-call cost engineering /liliFull-stack pragmatism: strong Python; enough TypeScript / Chrome-extension knowledge to own the integration /liliClear spoken and written English for demanding US executives /li /ulh3Knowledge /h3ulliContact-center / agent-assist patterns and metrics (handle time, cost per call, concurrency) /liliProduction LLM operations: load testing, data isolation, incident handling /liliNice to have: empathy-sensitive domains (healthcare, veterinary, insurance) and PE-sponsored rollouts /li /ulh3Experience /h3pKey characteristics (screen for all four): /pulliVoice AI in production — mandatory. Shipped at least one real-time voice or speech product to real users (agent assist, voice bot, live transcription copilot). Candidates will demo real artifacts at the interview. /lili6+ years hands-on AI/ML engineering, with strong recent LLM production practice /liliLatency and reliability record. Can show measured P95 reductions and concurrency fixes on a live system /liliConsulting / client-facing seniority. Calm and precise under detailed UAT scrutiny; manages expectations well /li /ulh3Nice to have: /h3ulliChrome extension delivery; telephony / streaming stacks (Amazon Connect, Twilio, LiveKit) /liliLangfuse in production /liliUS client experience with Eastern-time overlap /li /ul /p #J-18808-Ljbffr