Senior AI Engineer

Curie•San Francisco, CA

23d

About The Position

Curie is a telehealth platform that combines clinical expertise with AI to deliver personalized, accessible care. Our platform guides patients through intelligent intake, matches them with the right treatments, and supports clinicians with AI-powered tools — all built to be safe, compliant, and scalable. Our team includes ex-founders, clinicians, and engineers from leading institutions such as Stanford, Harvard, UCLA, Berkeley, and AWS. We're well-funded, growing aggressively, and building core infrastructure that will power the future of how patients access care. We're looking for a Senior AI Engineer to design and build the AI systems at the center of Curie's clinical platform. You'll own the Python and Go service layer that powers our clinical AI processing — from multi-step intake reasoning to retrieval-augmented generation for treatment guidance. This is an opportunity to shape the AI architecture of a healthcare product from the ground up, working closely with founders, clinicians, and engineers across the stack.

Requirements

7+ years of software engineering experience, with meaningful time building production AI/ML systems.
Strong Python expertise — you're comfortable with async services, and modern tooling (uv, Pyright, etc.).
Familiarity with PyTorch, TensorFlow, or Hugging Face Transformers for custom model work.
Hands-on experience with LLMs in production: prompt engineering, structured output, evaluation, and iteration — across commercial APIs (OpenAI, Anthropic, Google) or open-source models (LLaMA, Gemma, etc.).
Experience building RAG pipelines, vector search, or retrieval systems for grounding LLM outputs — using tools like LangChain, LlamaIndex, or custom implementations.
Familiarity with agentic AI patterns — multi-step reasoning, tool use, and orchestration frameworks (LangGraph, Google ADK, CrewAI, Claude Agent SDK, or equivalent).
Comfort working across service boundaries — you can navigate a Go backend, gRPC interfaces, and cloud infrastructure when needed.
Strong intuition for system design that balances correctness, observability, and performance.
Curiosity about healthcare and a desire to build AI that's safe, explainable, and clinically useful.

Nice To Haves

Experience with cloud ML platforms: GCP/Vertex AI, AWS SageMaker, or Azure ML.
Hands-on with local/self-hosted LLM inference: vLLM, Ollama, TGI, or GGUF-based deployments.
Fine-tuning or distillation experience — LoRA, QLoRA, RLHF, DPO, or similar techniques.
Familiarity with model evaluation frameworks (RAGAS, DeepEval, custom evals) and LLM observability tools (Langfuse, LangSmith, Arize, Weights & Biases).
Familiarity with healthcare data standards (FHIR, HL7) or EHR integrations.
Background in medical AI safety, bias detection, or clinical validation.
Experience with PostgreSQL (including JSONB, pgvector), sqlc, or gRPC/Connect-RPC.
Startup experience — especially as a founder, founding engineer, or early employee.
Published work or deep domain knowledge in healthcare AI or clinical NLP.

Responsibilities

Design and implement multi-step AI agent pipelines that process patient intake, synthesize medical history, and surface clinical recommendations.
Build orchestration patterns for managed, observable AI/ML workflows on cloud infrastructure.
Own session and memory management for long-running clinical agents — ensuring continuity, safety, and auditability across patient interactions.
Build and optimize RAG pipelines that ground clinical AI outputs in authoritative medical guidelines, drug references, and treatment protocols — including embedding models, vector stores, and reranking strategies.
Improve retrieval accuracy, citation traceability, and relevance ranking to ensure AI-surfaced information is trustworthy and explainable.
Continuously evaluate and iterate on retrieval quality with structured benchmarks.
Extend and maintain the Python service that communicates with other microservices, handling structured clinical data and LLM integrations.
Build data pipelines for ingesting, normalizing, and reconciling health data from external partners and EHR integrations.
Design systems that respect HIPAA requirements end-to-end — from data handling to model I/O to audit logging.
Instrument AI workflows with tracing, logging, and evaluation hooks for compliance-grade visibility into model behavior.
Build validation layers and guardrails that ensure clinical outputs meet safety thresholds before reaching patients or providers.
Monitor failure rates, latency, and model drift across production AI systems.