Staff AI Engineer (Hybrid)

Work Flexibility: Hybrid We're hiring a Staff AI Engineer to build GenAI and voice agents for medical devices, deployed both on-device and in the cloud. You'll own the technical direction for these systems — connecting clinical use cases to the models behind them (ASR, TTS, SLMs, speech-to-speech) while working within tight on-device limits la tency, memory, and reliability. This is a hands-on applied ML role, and the core challenge is making stochastic models behave predictably enough for clinical use: bounding them with deterministic architecture, building evaluation sets and frameworks, and designing safety guardrails that hold up in a regulated environment. At the staff level, you'll set the architecture and evaluation standards the rest of the team builds against, take on the hardest technical bets first, and align device software, data, validation, clinical, and regulatory teams around the safety, effectiveness, and quality of AI-enabled features across the product lifecycle — consistent with FDA guidance and good machine learning practice.

Own the technical direction of GenAI and multimodal agent (voice, text, vision) capabilities: translate product needs into robust, testable AI system designs, drive the architecture across components, and carry the highest-risk pieces from prototype through validation-ready implementation.   Architect stateful agentic systems (intent handling, tool/function calling, dialog and device-state management, interruption handling and recovery) that behave deterministically where safety requires it, with well-defined interface contracts between AI components and device software.   Manage and mitigate stochastic model behavior: design layered guardrails (deterministic validation, plausibility bounds, model-based checks), define what the model is and is not permitted to decide, and make those boundaries testable.   Set the standard for evaluation of m ultimodal GenAI systems: comprehensive test sets and automated harnesses covering task and intent accuracy, robustness under realistic clinical audio conditions, conversational quality, and responsiveness.   Develop and evaluate real-time speech and language components (speech recognition, synthesis, and dialog/turn handling), balancing model quality against the latency, memory, and reliability constraints of medical hardware.   Evaluate, select, integrate, and fine-tune off-the-shelf and small-footprint models (SLMs, domain-adapted ASR) for domain-specific terminology; own the buy/adapt/build decisions and their justification.   Establish safety, bias, and performance metrics tailored to voice and generative systems, and produce documentation supporting QMS and regulatory submissions.   Instrument systems for traceability: structured logging, audit trails of agent actions, and reproducible evaluation runs suitable for a regulated development process.   Mentor junior engineers, ensure engineering quality through design and code review, and communicate AI constraints and trade-offs clearly to product, clinical, and regulatory stakeholders.   Stay abreast of the rapidly evolving GenAI model, speech, and agent-architecture landscape; identify which advances matter for the roadmap and pragmatically incorporate them.   What You Need (Minimum Required Qualifications) Bachelor's Degree in Computer Science , Machine Learning, Electrical Engineering, Biomedical Engineering, Mathematics, or related field. 4+ years of AI/ML engineering experience, OR Master's Degree in the above fields and 2+ years.

(Strongly Desired) Strong proficiency in Python; optionally, working proficiency in C++ or other relevant languages for performance-critical and embedded integration work   Hands-on experience designing custom evaluation metrics, evaluation harnesses, and test-set generation ( both synthetic and human ) for stochastic AI systems.   Experience bringing statistical rigor to AI