How to apply Apply via the Linked In job opening, or send your CV and a short note to [Click to show email]
About the Project We are building motor-learning software that helps users learn physical skills using camera-based perception and AI. The system analyzes movement through computer vision and provides guidance during practice and feedback after practice. The LLM operates on structured outputs from vision systems (e.g., pose, skeleton, task state), not on raw images mostly. It serves as a controlled instruction and decision layer, where correctness and conservative behavior are critical.
Responsibilities Build LLM-based instruction and feedback logic using structured perception outputs (prompt schemas, constraints, and validation) Convert pose, task state, and confidence signals into concise verbal guidance Integrate LLM components into real-time pipelines Collaborate with CV/AI engineers on perception-to-instruction integration
Required Qualifications Strong experience deploying LLMs in production systems Experience working with structured inputs derived from vision systems Familiarity with schema-based generation, function calling, and guardrails Ability to reason about uncertainty and confidence explicitly Strong software engineering skills with focus on robustness
Strongly Preferred Experience with voice-based feedback or audio guidance systems Experience with on-device or latency-sensitive environments Background or interest in fitness, sports or human movement Experience combining rule-based logic with LLMsExperience with VLMs and Computer Vision in general.
How to apply Apply via the Linked In job opening, or send your CV and a short note to [Click to show email]