SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
BrightAI is a high-growth Physical AI company that processes visual, spatial, and temporal data from billions of real-world events across edge devices, mobile sensors, and cloud infrastructure to enable intelligent decision-making at scale.
You will own and evolve the on-device conversational AI powering an industrial safety wearable that guides field technicians through safety-critical procedures by voice. This is a systems engineering role, not a prompt-and-retrieve role. The work sits at the intersection of natural language understanding (NLU), small language models (SLM), and embedded software, where a wrong answer is a safety concern, not merely a quality issue.
Key responsibilities include:
- Own the on-device dialog pipeline end-to-end: intent routing, hybrid intent classification (pattern matching + embedding similarity + out-of-domain detection), text normalization for noisy speech input, and the multi-step guided-procedure engine.
- Maintain and extend the deterministic safety layer wrapping the language model—confirmation and echo-back gating, criticality tagging, negation handling—ensuring misheard answers on safety-critical steps cannot pass silently.
- Run SLM inference on-device under memory, computational, and latency budgets; reduce per-turn inference cost through model selection, quantization, and runtime optimization.
- Preserve and extend the zero-shot configuration model so new device commands and customer procedures are authored as data, enabling customer onboarding in hours rather than weeks.
- Coordinate the device deployment pipeline with the edge team and maintain the API contract with the on-device voice pipeline and speech-to-text (STT) stack.
- Define and run on-device benchmarks (latency, accuracy, false-accept/reject rates on safety-critical steps) and use measurements to drive engineering decisions.
- Build and maintain golden datasets and a non-regression suite as the command and procedure catalogs grow.
- Lead the migration from zero-shot to fine-tuned on-device models to reduce latency without reintroducing per-customer retraining burden.
- Collaborate with product, firmware, and cloud teams to bring new capabilities online, including additional languages, device commands, and guided workflows.
Required qualifications:
- 5+ years of experience in ML/AI with strong focus on NLP, LLMs, or conversational AI.
- Strong applied experience with LLMs: prompting, structured output, tool/function calling, evaluation, and RAG, with judgment to recognize when a model should not be used.
- Solid command of embeddings and semantic similarity (cosine similarity, centroid vs. maximum-similarity strategies, threshold tuning, out-of-domain detection).
- Strong Python with ability to write clean, tested, reviewable code; fluent with pytest, Git, and CI discipline.
- Strong experience building edge conversational systems, including multi-turn dialog/state management and efficient intent/NLU pipelines using local-first, cheap-to-expensive inference strategies.
- Experience with disambiguation, repair, and clarifying questions rather than committing to confident wrong answers.
- Comfort placing deterministic guardrails around probabilistic models, including safety floors, confirmation gating, negation handling, and state-machine or workflow engines.
- Experience running models on constrained hardware (NPUs, mobile, embedded targets) under real latency and memory budgets; familiarity with ONNX, onnxruntime, model quantization, and cross-architecture packaging for aarch64.
- Practical embedded development workflow experience.
