Jay Mehta

• Architected RAG-based knowledge graph with vector search and semantic memory, enabling AI agents to retain 10K+ user facts. • Designed multi-tenant OAuth 2.0 architecture supporting 5+ Gmail/Calendar accounts per user with automatic token rotation. • Built event-driven iMessage pipeline with async processing and HEIC conversion, handling 500+ daily attachments in real-time. • Developed distributed automation engine using BullMQ task queues and event routing, orchestrating 1,000+ daily AI workflows. • Deployed containerized services on Railway with GitOps, CI/CD pipelines, and observability achieving 99.9% uptime. • Designed PostgreSQL schemas with Row-Level Security, database triggers, and 10+ idempotent migrations for multi-tenant isolation.

• Architected hyper-personalized memory system using LangMem with semantic/episodic context, delivering tailored stories for 10K+ users. • Built production ready MCP clients enabling LLM-tool orchestration including a custom disposable environment using Pyodide. • Implemented multi-agent workflow using LangGraph with Kuzu graph database with optimized network headers to reduce latency. • Deployed Kubernetes-based infrastructure with Vessel containers, supporting daily agent interactions with minimal downtimes. • Developed observability stack using LangSmith evaluation metrics.

• Taught Analysis & Algorithms, focusing on time complexity, divide & conquer, and graph algorithms. • Developed a CI/CD pipeline using Python and Git for automating grading scripts and ensuring test coverage. • Received and provided feedback to enhance course delivery and student learning experience.

• Developed and trained models for fashion trend prediction, achieving 86% accuracy in using unsupervised learning. • Fine-tuned Yolov6 through Region of Interest (RoI) Pooling and image annotation on dataset using LabellImg.

• Cleaned and processed datasets from Kaggle using Pandas and NumPy and extracted leading KPIs to increase engagement by 25%. • Conveyed Data-Driven decisions to stakeholders by creating elaborate dashboards and visualizations using Tablaeu and Matplotlib.

• Developed and integrated Intelligent Chatbots deployed on Facebook and WhatsApp using the Jio Haptik platform, improving query response latency and distributing workloads efficiently to improve business revenue by $100,000+. • Assisted with the design, support, and development of new products using React and Nodejs. Developed APIs using Postman to retrieve and process data from Amazon S3 buckets, integrating with backend services on EC2.

• Utilized AutoCAD and Google Earth Pro to map geographical regions for a Watershed Management program in Maharashtra, in collaboration with IAHV. • Assisted in the conservation of rainwater in Maharashtra through the implementation of the program. • Contributed to the efforts of The Art of Living in promoting sustainable practices for water conservation in Mumbai.