Vansh

Architected and deployed Celery-based audio moderation and Kafka-based image & chat moderation pipelines using feature flags, sustaining 90%+ accuracy while moderating 500,000+ video bookings/day across production apps. Built a Django/Celery auto-retraining pipeline triggered when S3 frame growth crossed 100,000, with Slack notifications for automated retraining visibility. Designed acoustic analysis pipelines extracting silence duration and speaker counts via diarization for downstream content-quality scoring.

Backend Developer Intern at Lokal (2026-01 – Present)

• Architected and deployed Celery-based audio moderation and Kafka-based image & chat moderation pipelines using feature flags, sustaining 90%+ accuracy while moderating 500,000+ video bookings/day across production apps.
• Built a Django/Celery auto-retraining pipeline triggered when S3 frame growth crossed 100,000, with Slack notifications for automated retraining visibility.
• Designed acoustic analysis pipelines extracting silence duration and speaker counts via diarization for downstream content-quality scoring.
• Improved accuracy of AI moderation flagging system from 65 to 85+ percent

Full Stack Developer Intern at QuickIntell (2025-03 – 2025-12)

• Built a Python/Flask RAG system with LLaMA4-Maverick and Pinecone to digitize 30,000+ handwritten prescriptions, significantly improving data accuracy.
• Developed a Next.js virtual staging platform (oflisting.com), achieving a 95+ Lighthouse score and driving a 15% increase in user acquisition.
• Built QuickTrust (trust.quickintell.com) with Next.js/PostgreSQL — a compliance platform with a RAG engine for automated policy auditing.
• Contributed to QuickRCM (dev.quickintell.com), integrating AI services for medical coding, prior auth, and risk estimation.

B.Tech in Information Technology – Indian Institute of Information Technology Una (2022-08 – 2026-05)

Class XII – Tiny Tots Public School (Senior Secondary) (2022 – 2022)