Vaibhav Tanwar

Built a real-time multi-modal product retrieval engine using KD-Tree spatial indexing and inverted indices for O(1) categorical filtering, achieving <50μs query latency and 100+ QPS across thousands of fashion SKUs with thread-safe concurrent search infrastructure, validated by automated tests including latency regression baselines and concurrent stress tests. Architected and deployed an AI-powered fashion recommendation system on AWS EC2, integrating Adaptive SegFormer B2 for semantic segmentation with multi-modal CLIP embeddings (FashionViL, FashionCLIP, OpenAI models) and Weaviate vector search; added Redis-backed perceptual-hash caching that cut redundant ML inference by 40%+, adaptive sliding-window rate limiting, S3 presigned URL delivery with exponential-backoff retries, intelligent layout templates for outfit composition, and scaled RQ workers horizontally with GPU/CPU auto-detection. Containerized service using multi-stage Docker builds, monitored with CloudWatch metrics and Slack alerts, and wrote unit/integration tests achieving 85%+ coverage, resulting in sub-3-second outfit composition latency in production. Architected a production ML platform for ethnic-fashion insta

Engineered a scalable, Dockerized MLOps platform utilizing annotation enabled continual learning to enhance wildlife monitoring capabilities using camera trap images, achieving <60ms average API latency across all modules and scalable data handling of over 1M image records for a seamless Human-in-the-Loop annotation experience. Improved detection accuracy of multi-class endangered species by 54% by finetuning YOLOv8 on proprietary wildlife dataset from Wildlife Institute of India, and cut inference latency by 97% on average using TensorRT quantization, enabling near real-time processing. Built a scalable Open Set Re-Identification service using MegaDescriptor embeddings and CLIP model, leveraging PostgreSQL with pgvector, for efficient low-latency vector and semantic searches across millions of images. Designed and implemented Active Learning pipelines to alleviate high labelling cost in species segregation and Bird Count modules by selecting more informative samples to label based on instance level uncertainties, achieving a 3.5% increase in Mean Average Precision and 26.3% decrease in annotation budget.