AI Research Engineer – Generative AI

NeuroPace•Mountain View, CA

23h•Onsite

About The Position

Based in Mountain View, CA., NeuroPace is a commercial-stage medical device company focused on transforming the lives of people suffering from epilepsy by reducing or eliminating the occurrence of debilitating seizures. Its novel and differential RNS System is the first and only commercially available, brain-responsive platform that delivers personalized, real-time treatment at the seizure source. At NeuroPace, employees are our greatest asset. We are continually searching for solution-oriented individuals who can bring energy and creativity to our growing workforce. At NeuroPace, our success depends upon our ability to recruit and retain the most talented, enthusiastic and dedicated people we can find and providing them with a dynamic and challenging environment in which to thrive. We are currently seeking an AI Research Engineer to join our team to develop multimodal LLMs, reasoning agents, and large‑scale RAG systems . This individual will directly help people living with neurological disease; you’ll be part of a fast-moving group dedicated to using AI tools to accelerate impact of an FDA-approved implantable device . You’ll own model training pipelines, high‑throughput data and indexing systems, and retrieval‑augmented generation powering production applications. Ideal candidates have experience training large multimodal models, building scalable data/feature stores, and shipping agentic systems with strong observability and safety. Ideal candidates will be Bay

Requirements

2+ years in ML/AI (or equivalent impact).
Master’s in CS or related field, or equivalent experience.
Hands‑on LLM/multimodal training (PyTorch/TensorFlow/HuggingFace/SageMaker).
Strong engineering skills: Python, TypeScript, APIs, testing, profiling, and observability.

Nice To Haves

4+ years in ML/AI and production ML system experience.
Experience shipping agentic systems with planning, tool use, schemas, memory, and guardrails.
Multimodal expertise: CLIP‑style contrastive models, BLIP/LLaVA/Flamingo, Whisper/NeMo, video, neural signals.
Production RAG (chunking, grounding, hallucination mitigation, cost/latency tuning).
Distributed training (FSDP/DeepSpeed/ZeRO), mixed precision, checkpointing.
Information retrieval: query understanding, retrieval‑aware generation, knowledge graphs.
Causal inference or counterfactual evaluation for scientific/biomedical ML.

Responsibilities

Model Training & Inference Train/fine‑tune LLMs and multimodal models (text/vision/audio; bonus: neural/ECoG). Use SFT, DPO/ORPO, RLHF/RLAIF, and parameter‑efficient tuning (LoRA/QLoRA). Build agentic systems (tool use, planning, memory, reasoning) with attention to latency, safety, and cost. Optimize inference: quantization, graph/tensor optimizations, speculative decoding, distillation, caching.
Data & Indexing Build fast pipelines for ingest, feature extraction, multimodal alignment, and dataset curation. Design billion‑scale indexing (vector, hybrid sparse+dense, graph‑based). Implement observability: lineage, drift, coverage, bias, differential performance, cost.
RAG & Knowledge Systems Build production RAG (chunking, hybrid search, reranking, context optimization). Maintain continuous refresh pipelines (incremental indexing, TTL, schema evolution). Integrate ontologies/knowledge graphs for grounding and attribution.
Evaluation, Safety & Reliability Develop structured evaluations for faithfulness, factuality, consistency, and multimodal alignment. Implement guardrails: prompt hardening, policy filters, PII/PHI handling, output verification. Run A/B tests; define metrics for quality, latency, and cost.
Platform & Collaboration Partner with infra for distributed training (FSDP/ZeRO, multi‑GPU/TPU), registries, and ML CI/CD. Mentor engineers and contribute to long-term technical strategy.