AI/ML Engineer

This isn't a typical AI/ML role BRIIDGE is an AI-powered self-understanding platform. Users talk to Shirin, an AI companion grounded in 10 validated psychological dimensions - and every conversation builds a progressively accurate Living Profile. Mental health apps fail at 4% fifteen-day retention because they treat everysession as disposable. We are building the opposite of that. We are a team of 5. You will own the AI layer: the LLM orchestration, the profiling engine, the signals that power Shirin's memory and reasoning. This is production AI handling sensitive psychological data - it demands rigor, not just experimentation. What you'll work on You will architect and maintain the AI/ML systems that drive real-time conversation, multi-dimensional profiling, and crisis detection. You will integrate and orchestrate LLMs, design evaluation frameworks to measure model quality over time, and collaborate with the software engineering team to ship these capabilities as production-grade features in a TypeScript monorepo. Must have Hands-on experience building and deploying ML or LLM-powered systems in production Strong software engineering fundamentals - you write code that others can maintain Experience with LLM orchestration, prompt engineering, RAG, or fine-tuning Comfort working in TypeScript or strong willingness to adopt it as the primary language Genuine interest in building responsibly in the mental health and self-understanding space Nice to have MLOps experience: model versioning, monitoring, evaluation pipelines Familiarity with AWS and cloud-based ML infrastructure Experience with real-time streaming and event-driven architectures Background in healthcare, mental health tech, or therapy-adjacent products Agentic development workflow experience (Claude Code, Cursor, or similar) Knowledge of psychological frameworks or structured profiling methodologies What makes this different You are not joining a team running ML experiments that never reach users. We have a ratified architecture, deeply specified product requirements, and locked feature specs. You will understand why you are building what you are building. The product is real, the users are real, and the domain - human self-understanding - demands that the AI powering it be thoughtful, precise, and genuinely good.