At Joshnik Al Labs, I optimized a brand-intelligence agent serving 16K+ brands, reducing inference latency from 18 seconds to 4.2 seconds through parallel retrieval, deterministic ranking, context reduction, and MLflow evaluation.
I also architected a multimodal AI platform for a 6,000+ participant event, combining a DSPy/ReAct campaign agent with face search and WebSocket streaming. My work there also includes professional discovery across 50K+ profiles and production LLMOps for evaluation, tracing, and telemetry.
As a Machine Learning Intern at Alemeno Private Limited, I engineered a hybrid vision–LLM system for LCD defect detection across 17 classes, improving accuracy from 34% to 92%. I also built preprocessing and hard-negative mining pipelines and trained models on Vertex AI.
My projects include Curalink, an AI medical research assistant that placed third in an AI Hackathon, and PDF Corpus Chat, a multi-document RAG system. I’ve also built an agricultural RAG assistant and contributed to CNN-based skin disease classification pipelines.
