Where systems & open source meet. • Building, shipping, and mostly open-sourcing from Kakinada, India 🇮🇳
class MachaPraveen:
def __init__(self):
self.role = "AI Systems Engineer & Prompt Engineer"
self.founder = "ROLLNO31"
self.location = "Kakinada, India"
self.focus = ["AI agents", "MCP servers", "developer tools",
"medical imaging", "computer vision"]
self.mindset = "design it, ship it, open-source it"
def current_work(self):
return "building the quiet infrastructure between intelligent systems"- 🔭 I build AI systems that border on precision, intelligence, and open source — agents, MCP servers, and the tooling that connects them.
- 🩺 I ship real-world ML: glaucoma screening (U-Net segmentation), rebar detection, and other computer-vision pipelines — end to end.
- 🛠️ I love turning research into things people can actually run — mock modes, tests, clean schemas, the boring stuff that makes tools trustworthy.
- 🚀 Founder of ROLLNO31 — where systems & open source meet.
- 📫 Reach me at praveenmacha777@gmail.com or through my portfolio.
|
Full-stack deep-learning pipeline for glaucoma screening from fundus images — U-Net optic-disc/cup segmentation with a React + Flask app. Interactive crop → predict → graded result.
|
🛰️ lanforge-mcpAn MCP server letting Claude (or any MCP client) drive Candela's LANforge Wi-Fi test platform over JSON-over-HTTP. 6 tools, safe dry-run defaults, mock mode, 88% coverage.
|
|
Computer-vision pipeline that detects and counts rebar from construction imagery — selection, results overlay, and reporting.
|
🎓 ROLLNO31The venture I founded — open-source-first systems & developer tools. Where the ideas ship.
|
|
One-click highlight marking for long-form live streamers — Chrome extension + web app, Squad Sync multi-POV (every teammate's angle of the same second), editor handoff with Premiere / Resolve / Final Cut exports. My first product built for non-engineers + monetization.
|
- 📌 I treat prompting as engineering — structured system prompts, tool/function schemas, and evals — not guesswork.
- 🤖 I build agentic + MCP tooling so LLMs can safely drive real systems (see
lanforge-mcp: 6 tools, mock mode, 88% coverage). - ⚡ I ship faster with AI-assisted development across Claude, Cursor, and Copilot — from spec → code → tests → deploy.
- 🎯 Comfortable designing RAG pipelines, prompt chains, and structured-output workflows that stay reliable in production.
Whether it's collaborating on a project, a challenging problem, or just to talk tech — reach out.
⭐ Design it. Ship it. Open-source it. ⭐