Projects
Production ML systems, research projects, and open-source contributions showcasing expertise in NLP, Computer Vision, MLOps, and Edge AI.
Intent Detection & Query AutoComplete at Scale
Built real-time intent recognition system achieving <500ms latency with ~70% accuracy improvement. Implemented Query AutoComplete using Ghosting approach with N-gram models, custom vocabulary, and fine-tuned GPT-2 for intelligent suggestion ranking at SumoLogic.
Learn more
Enterprise MLOps Platform on AWS
Designed and deployed an automated MLOps system that reduced ML cycle time by 90%. Manages end-to-end model lifecycles: training, versioning, serving, A/B testing, and drift detection using AWS SageMaker, MLFlow, and Kubernetes.
Learn more
Edge AI & Model Optimization
Improved edge inference speed by 25-30% via quantization (QAT, TFLite, TensorRT, OpenVINO) and pruning. Built DeepOps for real-time edge device management—provisioning, tracking, health monitoring—reducing model management time by 80%.
Learn more
Computer Vision for Solar Construction
Reduced weekly construction site reports time by 70% using ML models (YOLOv8, UNet, SAM, Transformers) for materials detection. Built event-driven integration handling 1000+ concurrent jobs with FastAPI, Airflow, and AWS Lambda.
Learn more
MAVI: Mobility Assistance for Visually Impaired
Led a team of 6 at IIT Delhi developing intelligent edge devices for visually impaired individuals. Implemented object detection, signboard reading, and face recognition achieving 40-70% faster processing on Raspberry Pi + Neural Compute Stick 2.
Learn more
Detecting Harmful Content on Instagram
System design exploration on detecting harmful content using NLP, Computer Vision, and ML at scale. Covers sentiment analysis, hate speech detection, content filtering, and real-time moderation strategies.
Learn more