An OpenEnv environment where a vision-language agent audits and repairs deliberately corrupted annotations on real COCO val2017 images. Sidesteps the usual failure mode of VLM bounding-box regression by rendering ID tags directly onto the image (a Set-of-Mark overlay), which turns the model into a pure semantic auditor instead of asking it to hallucinate coordinates. Three difficulty tiers (remove spurious, fix classes, find missing) with dense per-step rewards, anti-loop penalties, and deterministic 0–1 grading.

Engineered a DistilBERT safety classifier (87% F1 across 3 classes) with batch inference and confidence scoring, plus an end-to-end red-teaming pipeline (500+ adversarial prompts, 3 harm categories, 3 mitigation agents) that reduced unsafe generations by 42%. Built an observability suite - confusion matrices, per-class metrics, and JSON/visual exports - for real-time monitoring and drift detection in production LLM safety workflows.

Built for Adobe DevCraft. Innovated a raw-to-dense feature pipeline that transforms raw bid data using contrastive embeddings and autoencoders, followed by ANN-based feature compression for efficient representation. Devised a low-latency RTB model predicting bid price and bidding decisions within 5ms per request, tuned via grid search and adaptive learning-rate scheduling. Achieved 82% classification accuracy, scaling to 100K+ bid requests per second in large-scale ad exchanges.

An ML system that identifies distinct market states from high-frequency financial data. Engineered 20+ custom features from order-book data and consolidated 3 clustering algorithms to an 87% silhouette score. Identified 4 distinct market regimes at 92% classification accuracy, reducing strategy drawdowns by 15% through adaptive position sizing. Analyzed 500,000+ data points across multiple timeframes for real-time regime classification within 50ms.

Translating 2-band SAR imagery into 13-band multispectral EO images using CycleGAN while preserving semantic and spectral consistency across domains. Focused on stabilizing adversarial training and faithful spectral reconstruction for remote-sensing data.

Built a real-time, cross-platform AI search assistant with Flutter and FastAPI. Created a custom ranking algorithm using sentence transformers achieving 85% accuracy in source ranking, with WebSocket streaming for live responses and a RAG pipeline combining the Gemini API with Tavily Search for accurate information synthesis.

Collaborated on a mental wellness companion application using the MERN stack, recognized as a top project by Google Developers Club Delhi. Designed responsive UI with React and TailwindCSS, integrated Firebase analytics and MongoDB for secure data management. Built stress analysis using sentiment analysis and mouse tracking at 92.71% accuracy.