- Company Name
- OutcomesAI
- Job Title
- Tech Lead — ASR / TTS / Speech LLM (IC + Mentor)
- Job Description
-
**Job title:** Tech Lead – ASR / TTS / Speech LLM (IC + Mentor)
**Role Summary:**
Lead end-to-end development of ASR, TTS, and Speech‑LLM models for a medical AI platform. Drive architecture design, training strategy, benchmarking, and production deployment while mentoring a small engineering team. Ensure models meet performance, latency, and clinical accuracy standards in a HIPAA‑compliant environment.
**Expectations:**
• Own technical roadmap for speech model lifecycle
• Mentor and guide engineers on data curation, fine‑tuning, and serving best practices
• Deliver production‑ready speech models that meet clinical benchmarks (e.g., WER, entity F1, latency)
• Collaborate with backend/ML‑ops for observability, health metrics, and autoscaling
**Key Responsibilities:**
1. Define and execute training pipelines (LoRA/adapters, RNN‑T/CTC, multi‑objective losses).
2. Evaluate and benchmark open‑source models (Parakeet, Whisper, etc.) against internal test sets.
3. Design synthetic data generation and augmentation pipelines (speaker selection, noise, codec simulation).
4. Architect inference stack: Triton Inference Server, TensorRT/FP16, K8s autoscaling for 1,000+ concurrent streams.
5. Implement language‑model biasing APIs, WFST grammars, and context biasing for domain accuracy.
6. Lead evaluation cycles, drift monitoring, model switching, and fail‑over strategies.
7. Mentor engineers on best practices and conduct code reviews.
8. Partner with backend and ML‑ops teams to ensure production readiness and observability.
**Required Skills:**
- Deep expertise in ASR, TTS, and Speech‑LLM with PyTorch, NeMo, ESPnet, Fairseq.
- Proficiency in streaming RNN‑T/CTC, LoRA/adapters, TensorRT optimization.
- Hands‑on with telephony robustness (G.711 μ‑law, Opus, codec augmentation, AGC, band‑limit).
- Experience with speaker diarization, turn detection, and smart voice activity detection.
- Knowledge of TTS systems (VITS, FastPitch, Glow‑TTS, StyleTTS2, CosyVoice, BigVGAN).
- Proven ability in deploying with Triton, Kubernetes, and GPU scaling.
- Familiarity with LM biasing, WFST grammars, and context injection.
- Strong mentorship, code‑review discipline, and collaborative mindset.
**Required Education & Certifications:**
- M.S. or Ph.D. in Computer Science, Speech Processing, or related field.
- 7–10 years of applied machine‑learning experience (≥3 years in speech or multimodal AI).
- Track record of shipping production ASR/TTS models or inference systems at scale.