Sanyam Garg
Age:
Madison, WI
UW–Madison junior studying Computer Science & Data Science. I build infrastructure, backend systems, and tools that solve problems I actually have.
Education
University of Wisconsin–Madison
Aug 2023 - Dec 2026
Degrees
Relevant Coursework
Tools
Programming Languages
Python, Go, C, Java, SQL, Bash
DevOps & Infrastructure
Git, Docker, Tailscale, Linux, Cloudflare Tunnels, Flask, Node.js, Firebase, REST APIs, WebSocket, CI/CD, Containerization, GCP
Experiences
WiNGS Lab, Prof. Suman Bannerjee (UW–Madison)
Madison, WI | January 2026 - Present
Undergraduate Research Assistant
Got picked up by the Wireless and Networked Systems Lab to work on edge computing. Basically, making devices smarter without relying on the cloud. The work is focused on indoor localization for HoloLens 2 headsets using Monte Carlo Localization on building floor plans. The idea is to help first responders and people in high-stakes situations map out a building in real time, know where they are, and stay connected with each other through their headsets.
Century Financial
Dubai, UAE | June 2025 - Aug 2025
AI & Backend Development Intern
Built two major things here. The first was an AI Trading Assistant. I wrote the Python backend that connected OpenAI's API to a RAG system pulling live portfolio data, so users could ask the AI questions about their actual holdings. Used the Model Context Protocol (MCP) to orchestrate tool calls so the AI could securely read data and persist context across sessions.
The second was a real-time co-browsing engine. Instead of screen-sharing over video (slow, expensive), I built it with Node.js and WebSockets using DOM mirroring, which uses way less bandwidth. I also added client-side masking to redact sensitive financial data before it ever hit the network. That feature replaced a third-party tool and saved the firm $20,000+/year.
Projects
AirPipe – CLI File Transfer Tool
GitHubI needed to move files between my servers and my phone without installing anything. So I built it. One command, one QR code, done. Written in Go, it spins up a relay that your device connects to by just scanning.
The fun part: the relay is zero-knowledge. Encryption keys live in the URL fragment (never sent to the server), and everything is encrypted with NaCl secretbox (XSalsa20-Poly1305). The relay only ever touches encrypted bytes. Open sourced it, got 50+ GitHub stars, and it was featured on r/selfhosted.
I exposed my server's SSH to the internet just to see what would happen. Within minutes, bots from all over the world were hammering it with credentials like "root:admin". So naturally, I built a dashboard to watch it happen live.
The backend is a Python (FastAPI) pipeline that tails system logs over SSH, extracts attacker IPs, geolocates them, and serves it all through a REST API that the frontend polls every 1.5 seconds. Deployed with Docker and Cloudflare Tunnels. The map uses Leaflet.js to animate each attack as it lands. It's weirdly mesmerizing.
Private Cloud & Server
Network MapI didn't want to rent someone else's computer for everything, so I built my own cloud. It's a hybrid-cloud setup with cloud VPS + bare-metal Linux boxes in two countries, stitched together with a Tailscale mesh network.
Currently running 20+ services in Docker: media automation, dev tools, monitoring, you name it. Public access goes through Cloudflare Tunnels so my IP stays hidden. Wrote my own monitoring in Python and automated daily Bash/Cron backups to an off-site location. It's basically my playground for learning infrastructure by actually running it.