Full-stack AI engineer

## About the role We're building the first precision mental wellness platform, making thought patterns, limiting beliefs, and emotional regulation trackable, predictive, and transformable. ## The opportunity You'll be an AI engineer on a platform that: - Maps cognitive patterns through sentiment and voice analysis, detecting impostor syndrome, people-pleasing, perfectionism, and 30+ other patterns before they spiral - Predicts pattern activation 30-60 minutes in advance by fusing wearable data (HRV, sleep, movement) with voice markers and calendar context - Delivers precision interventions through conversational AI that adapts to each user's cognitive profile, learning style, and emotional capacity - Creates "Cognitive Twins", digital models that simulate users' thought patterns and predict long-term trajectories ## What you'll build ### Year 1: Pattern Recognition & Prediction Engine - Voice-based pattern detection: Build NLP models that identify cognitive distortions (catastrophizing, black-and-white thinking, personalization) from natural speech - Predictive analytics (v1.0): Create ML models that predict pattern activation using voice markers + wearable data + calendar context with 60%+ accuracy - Cognitive Twin architecture: Design the digital twin system that models users' belief structures, triggers, and behavioral responses - Real-time intervention optimization: A/B test which interventions work for which cognitive profiles, building a continuously learning system ### Year 2: Multimodal AI & Physiological Integration - Wearable data fusion: Integrate Oura, Apple Watch, Whoop data to correlate HRV, sleep, movement with pattern intensity - Emotion detection: Build models that detect emotional state from voice tone, pace, linguistic markers - Longitudinal trajectory modeling: Predict 6-month, 12-month outcomes based on current patterns (depression risk, burnout probability) - Improve prediction accuracy: 60% → 75% → 85% through continuous model refinement ## What you'll own - ML model development: Design, train, and deploy models for pattern recognition, prediction, and intervention optimization - Production ML infrastructure: Build reliable, scalable systems for real-time inference (voice AI, predictive analytics, recommendation engines) - Data pipelines: Architect ETL for voice transcripts, wearable data, user interactions, clinical outcomes - Model monitoring & iteration: Track model performance, identify drift, implement continuous improvements - Cross-functional collaboration: Work with clinical team (Dr. Adrienne Heinz from Stanford NCPTSD), product, and engineering to translate mental health science into production AI ## What you bring ### Must-Haves: - 5+ years productionizing ML models in real-world applications - Expert-level Python and ML frameworks (PyTorch, TensorFlow, or JAX—we're framework-agnostic) - NLP experience: Building models that understand human language (transformers, LLMs, speech recognition) - Production ML systems: Experience with model serving, monitoring, A/B testing, continuous training - Ownership mindset: You've taken models from research → production → measurement → iteration