Things I’ve built, from production platforms to weekend projects.

The work I’m most proud of: a platform I rebuilt end-to-end as the only engineer, an open-source job-tracking tool with a Claude-powered agent, and the smaller systems I build to learn. If something interests you, the code and the live products are linked.

  • JobWatch

    Open-source job-tracking platform (React 19, Firebase Gen 2) that monitors 950+ companies across nine ATS platforms including Greenhouse, Ashby, Workday, Oracle Recruiting Cloud, and Eightfold. A Claude-powered agent and custom MCP server score jobs and draft cover letters, with 80%+ lower LLM cost through Anthropic prompt caching. Hybrid rule-plus-LLM scoring and aggregation documents cut Firestore reads by 99%, and a browser extension auto-fills Greenhouse and Ashby applications.

    github.com/akash-ramasani/job-watch

  • RunnSmart platform rebuild

    Production rebuild of an AI-assisted advertising platform at Runn Technologies, where I am the sole engineer. Eight TypeScript services: a Fastify API with 200+ routes, a Temporal worker running 14 workflows, an MCP server, a headless page renderer, and Next.js customer, admin, and marketing apps, on a 79-table PostgreSQL schema with 1,300+ automated tests. Integrates 16 providers including Google Ads, Meta, Microsoft Ads, TikTok, Stripe, OpenAI, and Anthropic. Serves 1,200+ customers with 3,000+ campaigns launched.

    runnsmart.com

  • RapidRxNow Operations Hub

    Internal catalog-management application for pharmacy operations built with Next.js App Router, Prisma ORM, and PostgreSQL. Ingests POS datasets with batch processing and multi-store synchronization, streams progress to the browser over Server-Sent Events, and enforces role-based access with custom Edge Middleware and NextAuth.

    rapidrxnow.com

  • Kubernetes homelab and log monitoring

    Three-node Kubernetes cluster on repurposed hardware running Ubuntu Server, with a FastAPI service that exposes Linux logs through REST APIs instead of manual SSH. Reverse chunked reads scan multi-GB log files without loading them into memory, covered by pytest.

    github.com/akash-ramasani