About the Company/Team

RecoveryAI is pioneering the future of substance-use recovery through proactive, adaptive artificial intelligence. Our flagship product, Ebby™, is a 24/7 AI companion that continuously learns from daily behavior, mood, and context—delivering just-in-time adaptive interventions (JITAI) that provide the right support at precisely the right moment. Rooted in clinical research and authentic lived experience, Ebby™ represents a shift away from reactive tracking to proactive intervention, helping people stay engaged in recovery and reducing the risk of relapse. Our mission is to increase long-term sobriety by reducing relapse through predictive AI, real-time behavioral support, and interventions that adapt to each person’s unique recovery journey.

About the Role

We’re looking for a short-term, paid advisor (1–2 hours to start) with hands-on experience fine-tuning and evaluating a self-hosted LLM. The goal is to sanity-check our end-to-end approach: training data preparation, test case design, evaluation methodology, and what “success” should look like in production. Topics we want to cover: Fine-tuning approach for a self-hosted model (e.g., PEFT/LoRA, dataset sizing, formatting) Training data strategy (what to collect/label, how to avoid noise and leakage) Test cases + eval harness (expected outcomes, regression tests, safety checks) Practical go/no-go criteria and deployment considerations

This is a paid, short engagement suitable for advanced PhD students, postdocs, or experienced practitioners who have shipped fine-tunes in real systems. We value clear, decision-grade guidance over theoretical exploration. If there’s mutual fit, there may be an option for occasional follow-on advising, but the initial ask is intentionally lightweight.