Six years building backend systems, analytics pipelines and GenAI features at Oracle, Synopsys and Walmart Global Tech — and shipping Taskyard, a live AI-native task platform, on my own time.
Oracle Guided Learning — full-stack analytics used across Oracle Fusion applications
Develop features across the full stack in Python and React, contributing to Oracle Guided Learning's analytics capabilities across Oracle Fusion applications.
Build and extend ETL and data-processing workflows over analytics tables and OCI Object Storage buckets to support data-driven application features.
Develop GenAI capabilities on OCI Generative AI and Generative AI Agents, integrating AI-powered functionality into application workflows.
Build RAG pipelines and agentic workflows with LangChain and LangGraph for retrieval-grounded GenAI capabilities.
Develop AI-assisted workflows for guide creation and application automation, reducing manual effort.
Work on realm and tenant-specific AI model provisioning, supporting model configuration across Fusion environments.
Improve backend processing through asynchronous Python (asyncio) and system-level optimisation.
Resolved 50+ production and support issues, including vulnerabilities identified through OWASP scans.
Pillar-wide Oracle hackathon winner. Built an AI-powered analytics solution, later productized into a custom dashboard and automated chart-generation capability.
Rapid Response Team Award. Recognised for critical production support and fast resolution of high-priority issues.
Led the redesign and modernisation of a software suite for design, verification and release of IP/MSIP solutions, used by internal and external customers.
Rebuilt the frontend in React and Redux, cutting page-load times by 40%, and optimised PostgreSQL queries, indexing and caching underneath it.
Cut troubleshooting time by 30% and lifted customer satisfaction by 25% by resolving complex production issues.
Earned two internal awards and an early promotion for sustained performance improvements.
A full-stack, multi-tenant task and project management SaaS with subscription billing, built on Prisma and serverless Postgres (Neon), and an independent async worker queue for scheduling — designed to scale past simple polling.
Reminders run through Redis rather than polling Postgres: a scheduler process computes next-occurrence timing and enqueues jobs, while a separate worker consumes them in cursor-based batches with bounded concurrency.
Built for reliability under load — retry limits, queue-safety guards, and graceful shutdown on SIGTERM/SIGINT so in-flight jobs aren't lost on deploy.
Repositioning it as an AI companion for teams: tasks generated automatically from meeting transcripts, messages and check-ins, with MCP server support and agentic workflows underneath.
Pursuing distribution through MCP registries and directories, including the Official MCP Registry, Smithery, PulseMCP, Glama and mcp.so.
A dead-simple CLI checkpoint and rollback tool for EDA projects. Snapshots multi-gigabyte simulation workspaces in one command and restores them in seconds, no recompute needed.
Content-addressed, SHA-256 hashed object store designed to scale to multi-gigabyte EDA workspaces without duplicating unchanged files across checkpoints.
Ships as prebuilt binaries for Linux, macOS and Windows via GitHub Actions, alongside tagging, diff, export-to-tar.gz and garbage collection.
Built-in ignore rules for Vivado, Quartus and ModelSim byproducts, drawing on EDA workflow experience from Synopsys.
