Vedant Lavale

Vedant LavaleApr 17, 2026

Secured Top 5 at Ignition Hackverse hosted by Pune Vidhyarthi Griha's College of Engineering and Technology Pune . We built Veridium, an intelligent insurance claim processing system designed to enable Straight Through Processing for low complexity claims. This problem statement was proposed by Persistent Systems . The goal was to reduce manual intervention and allow claims to be evaluated and settled end to end in a fast and reliable way. In the current insurance flow, even simple claims go through multiple manual checks. This leads to delays, higher costs, and a poor user experience. With Veridium, we focused on automating these repetitive workflows while maintaining accuracy and fraud detection. We approached the problem in two ways. First, we built a traditional machine learning pipeline. This handled structured validation such as checking if the policy is active, verifying required fields, and detecting duplicate claims. We also extracted features using computer vision and text processing. For car claims, we used YOLO to detect damage and estimate confidence. For travel claims, we used OCR to extract key details from documents. Second, we explored an AI driven pipeline using OpenClaw. This helped us bring in more flexible reasoning and decision making on top of the structured pipeline. We then combined both approaches into a hybrid system. The system calculates multiple scores: - Consistency score between user input and model output   - Evidence score based on confidence from YOLO and OCR   - Policy score based on coverage matching using a retrieval system  All of these feed into a final scoring engine that decides: - Auto approve for high confidence claims   - Manual review for edge cases   - Reject for low confidence or invalid claims  Building Veridium pushed us to think beyond just coding features. We focused on system design, real world constraints, and how to balance automation with reliability. Big thanks to the organizers and management team at PVG College for putting together a well structured and high energy hackathon. The execution, problem statements, and overall experience made it genuinely worthwhile. Special thanks to our mentor Chinmay Relkar for suggesting impactful changes that helped us refine both the approach and the final system. Built this alongside Kanaad Limaye, Sarwadnya Mahajan, VISHWAJEET MORE. Strong team effort throughout. #Hackathon #MachineLearning #AI #SystemDesign #InsuranceTech #ComputerVision #OCR #Automation #Openclaw #Developers #OpenToWork

Vedant LavaleJan 28, 2026

Happy to share that our project won 1st place at the GDG WOW Verse Hackathon, organized by GDSC MMCOE, Pune 🚀 We built a VR-based physiotherapy rehabilitation platform that brings real-time posture validation, adherence tracking, and gamification to at-home recovery. Using computer vision–based pose detection and exercise-driven VR games, patient movements directly control gameplay incorrect posture blocks progress. A doctor-controlled backend enables exercise assignment, accuracy thresholds, progress tracking, and leaderboards for motivation. Built as Team Pre Trained Users with Chaitanya Luniya. We’ve qualified for the next round of the Google Hackathon and are excited to push this further. #Hackathon #GDG #VR #HealthTech #ComputerVision #Gamification 🚀

Vedant LavaleMay 21, 2026

what in the world is this URL structure Blinkit ?!