I build production-grade AI systems that ship!!
From cattle health computer vision to portfolio analytics to chatbots, I turn notebooks into shipped products
GauSwastha
An AI-powered computer vision platform for livestock health. Using posture, coat quality, and body condition scoring, and many other parameters, it detects early illness and estimates weight, breed, market value, milk yield and many other parameters directly from images. Deployed in production with the Government of Karnataka, GauSwastha runs on serverless infrastructure with cloud-based retraining, enabling large-scale animal health monitoring that is both scalable and cost-efficient.
Computer Vision
Serverless Inference
Active Learning
User-Friendly Interface
Cross-Team Collaboration
OptiRice
I developed an automated quality-assessment pipeline that scores rice grains on whiteness, broken ratio, and impurity levels using semantic segmentation and regression heads. Operators upload batch samples through a web interface and receive live dashboards with traceable metrics.
Semantic Segmentation
Quality Metrics
Enhanced Security
Centralized Control
Clustering
Markowitz Recommendation Engine
I built a microservice that ingests live market data, runs constraint-aware mean-variance optimization, and exposes portfolio suggestions via a REST API. A dynamic dashboard visualizes VAMI, drawdown alerts, and Sharpe/Sortino ratios, empowering advisors and retail users to make data-driven allocation decisions. The system runs reliably under load, delivering personalized recommendations to 500+ monthly users.
Financial Analytics
Trend Detection
Opportunity Identification
Enhanced Agility
Comprehensive Analysis
Orbis
I built a clustering-and-routing engine that handles real-world delivery constraints—shift windows, vehicle capacity, cash-in-hand, stop times, lunch breaks, and depot returns. Using a combination of DBSCAN, K-means, and Google OR-Tools, it plans 50+ stops in under 2 seconds and runs as a FastAPI microservice.
Crop Recommendation Engine
I built this engine in collaboration with agricultural scientists. The system asks farmers a few simple questions, such as what crop was sown before, what fertilizers they used, land size, and irrigation setup. It also automatically detects location from the mobile phone and factors in bedrock conditions, rainfall history, weather forecasts, and price trend predictions to recommend the most suitable crops for that farmer.
Voice-First Agri Chatbots
I developed multilingual voice assistants (English, Hindi, Kannada) using FastAPI and LangChain with intelligent intent parsing and context resolution. The system features a modular tool registry architecture that dynamically handles weather queries and agricultural services through structured LLM interactions.
