Company Description
Yuna's mission is to radically transform the way mental health support is accessed and delivered. By providing 24/7 instant and private support at the push of a button, we aim to address the significant gaps in mental health care accessibility. In a system where waiting times for therapists can reach up to 48 days and many people can't even afford care, Yuna is the answer to immediate mental health support needs.
We’re growing quickly, and every person who joins us plays a pivotal role in shaping the future of mental health technology.
Position Overview
You’ll own Yuna’s core conversational intelligence (i.e. spanning retrieval, memory, safety, evaluation, and post-training). You’ll ship user-visible improvements quickly while raising the bar on empathy, reliability, and guardrails. This is a hands-on staff-level role: equal parts research-to-production and platform thinking.
What you’ll do
* Lead the end-to-end model lifecycle of Yuna’s conversational AI: problem framing → data curation → modeling/fine-tuning/post-training → evals → rollout → monitoring
* Work closely with clinical psychologists (both internal and external to Yuna), ensuring the efficacy of Yuna’s conversational AI
* Design agentic flows (tool use, planning, multi-turn memory) and safe retrieval for mental-health use cases
* Build and own the evaluation flywheel: benchmarks, judge models, rubrics for conversational quality, ability to help, safety
* Implement alignment & safety layers (e.g., self-harm detection, red-teaming, refusal strategies, escalation paths) tuned to clinical guidelines
* Optimize for latency, cost, and quality
* Partner tightly with product, sales, and users to scope new features and
* Establish LLMOps: datasets, prompts/checkpoints/versioning, observability, drift monitoring, and incident response for model quality
What you’ll bring
* 5+ years in ML/AI with 3+ shipping LLM/NLP systems in production
* Deep skill in Python and modern deep learning stacks (PyTorch/JAX/TensorFlow), retrieval/RAG, embeddings, reranking, and long-context summarization
* Experience building agentic systems (tool calling, decomposition, memory) and rigorous eval harnesses (offline + online)
* Strong product sense; you care about user outcomes, not just model metrics
* Comfort working with ambiguity and creating systems 0 to 1
Location: Remote (work from anywhere, with at least 6 hours overlap with PST)
Employment Type: Full Time
What We Offer
* Competitive salary (based on experience) + equity options
* Remote-first culture with flexibility
* A fast-growing, talented, and empathetic team dedicated to transforming mental health care
* The opportunity to make a real impact on the world, building cutting-edge AI systems that improve lives every day