SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 113,800 - 154,525 / annual
Lumeris is seeking a Senior Data Scientist to build voice AI and model capabilities for Tom, an AI-enabled primary care platform. You'll report directly to the VP of AI and split your time between developing and fine-tuning clinical AI models and building voice systems for testing and validation before patient deployment.
Key responsibilities include building and deploying real-time conversational voice agents across the full speech stack (speech-to-text, LLM reasoning, text-to-speech, and speech-to-speech). You'll solve practical production challenges including telephony audio fidelity, latency, interruptions, background noise, and variable capture conditions. You'll develop simulation capabilities to generate and role-play patient conversations at scale, covering edge cases that manual testing cannot reach, and make architecture tradeoffs balancing control, latency, and simplicity.
On the model development side, you'll design and train deep neural networks from scratch, fine-tune language models on clinical and longitudinal healthcare data for prediction and generation tasks, and build data pipelines that make clinical data usable for model development. You'll select the right training approach for each problem and articulate where training, fine-tuning, and prompting belong.
You'll establish evaluation pipelines measuring model output for accuracy, consistency, completeness, and omission using LLM-as-judge approaches. You'll red-team AI agents to surface failure modes before production and translate findings into measurable quality improvements. You'll partner with clinical, product, and engineering teams to turn ambiguous problems into defined technical work, communicate complex concepts to technical and non-technical audiences, and mentor junior team members.
Required: Bachelor's degree in quantitative field (statistics, mathematics, economics, actuarial science) or equivalent; 5+ years in data science, machine learning, or applied AI; hands-on production experience building and deploying voice or conversational AI systems with real audio in live environments (not text-only NLP); demonstrated experience training neural networks from scratch and fine-tuning language models; proven track record shipping and hardening AI systems in production; strong Python programming and modern deep learning frameworks (PyTorch, TensorFlow); working knowledge of real-time audio infrastructure and telephony integration.
Preferred: Master's or PhD in quantitative/scientific discipline; GCP experience; healthcare or regulated-domain experience with clinical data standards (FHIR); voice model or conversational agent experience at AI labs or voice-first platforms.
