Hi, I'm Kavish Shah

Engineered distributed vector search infrastructure with HNSW indexing and cosine similarity algorithms, optimizing retrieval latency to sub-100ms and scaling to 10,000+ embeddings with 99.9% uptime. Built AI-powered document extraction system using Claude API and NLP pipelines, implementing Redis caching layer that cut processing time by 75%. Architected automated weekly digest system using Resend API and cron scheduling, implementing AI-driven content summaries and personalized email generation for user sustainability insights.

Implemented an AI-powered pipeline leveraging Random Forest and Gradient Boosting that achieved 96% predictive accuracy in soil moisture estimation, improving irrigation efficiency across diverse soil types. Reduced irrigation error variance by 40% by implementing soil-texture-specific calibration models and integrating Osmo's 143× larger sensing volume, resulting in more accurate root-zone water availability insights. Accomplished competitive accuracy within 5% of ensemble models by implementing Linear Regression, which decreased training complexity and computation time by 60%, informing cost-efficient deployment decisions.

Deployed scalable full-stack web applications using Next.js, PostgreSQL, and Redis; configured GitHub Actions CI/CD and Playwright end-to-end tests, reducing deal tracking efforts by 25%. Deployed a RAG chatbot implementing persistent long-term memory and resolving vector store/state management issues to reduce internal deal tracking time by 30% and improve multi-turn accuracy by 40%. Resolved critical memory persistence issues in multi-session AI workflows, increasing chatbot context retention by 80% and reducing duplicate query errors by over 50%.