Vedant Madane

Working with distributed US-based teams on cutting-edge AI/ML projects, collaborating asynchronously across 12+ hour timezone difference.

• Agentic Trajectory Annotation: Annotating and validating LLM agent trajectories for training AI systems that solve complex Kaggle-style ML tasks through iterative analysis, coding, and debugging
• Trajectory Reviewer Tool: Built a full-stack web application (React + Flask) for reviewing SWE-bench agentic trajectories with GPT-4.1 integration for AI-powered analysis and patch leakage detection
• ML Competition Solutions: Developed solutions for various Kaggle competitions including ASL signs recognition, aerial cactus identification, Otto recommender systems, and molecular translation challenges
• Remote Collaboration: Coordinating with US-based ML engineers, participating in async standups, and delivering quality work with minimal supervision across timezones

Worked in a hybrid remote environment, collaborating with distributed teams across Maharashtra's learning centers. Promoted based on performance and contributions.

• Created photo & video capture functionality with client-side posture detection, gender determination, and text extraction
• Built TensorFlow models to predict footfalls to learning centers
• Generated automated MAU reports using Pandas for stakeholder visibility
• Applied behavioral psychology insights to improve user retention

Complete immersion experience in Russia's two largest cities. Developed cultural adaptability and cross-cultural communication skills that translate directly to remote international work.

• Learned Russian through complete immersion into the culture
• Later worked as a freelance Russian interpreter for business negotiations
• Developed comfort with ambiguity and unfamiliar environments