Hari Varsha

Hi! I'm Hari Varsha, better known as vxsha-256 or nanodijkstra around the internet. From an early age, I’ve been captivated by computers, spending almost all my free time playing Pokémon MMOs, watching old-school noir films, and doomscrolling through internet forums learning random trivia in the process.

I'm a fresh Electrical and Computer Engineering grad. I've hacked around everything fun from large-language models to embedded microcontrollers and worked as a Summer Research Intern @ UCSC (Go Slugs!) and as a lab member for over a year. I specialize in building the next-generation serverless platforms and machine learning systems.

Over the years, I’ve become deeply entrenched in the world of deep systems programming, AI research, game engine development, and cloud infrastructure. Apart from intense development, I like consuming comics like XKCD and SMBC, reading Hacker News threads, and essays from cool people like Paul Graham & Joel Spolsky.

I lead the Arcane Systems Reading Group, a collective of systems nerds who explore niche topics like compilers, operating systems, distributed systems, low-latency systems, and performance engineering. My career goals can essentially be boiled down to producing the same influence as YEAT, DVRST, Playboi Carti, and Kendrick Lamar but for computer science as a whole.

Current

🚨 I am currently applying to graduate programs for computer science for Fall 2025, and I'm open to MS/PhD programs. If you'd like a hardworking & passionate student on your team, please reach out to any of my contacts found in the bottom of this website. Thank you so much for your consideration ! 🚨

I'm having a great time reading source code of Unix-based Kernels and dissecting Kernel-Level anti-cheats, implementing research papers like the Raft protocol and watching conferences such as CPPcon and Zig Showtime. I plan to set up substack and start streaming development on Twitch soon in the future. My hobbies include amateur game development, reading RFCs and learning Japanese.

I've also been deep diving into internals of Zig programming language, writing my own Linux Kernel Module in C and Zig, and experimenting on Hash Tables to learn them on a deeper level. I'm trying to build a habit of reading technical books and I've been enjoying Designing Data-Intensive Applications by Martin Kleppmann & Algorithm Design by Éva Tardos.

Ongoing Projects

My projects reflect my interests in machine learning infrastructure, hardware engineering, video-game development, and systems-level programming. Here are some projects I’m currently working on:

  • Memspect [C, Rust, LLVM]: A cool static analysis framework for real-world C codebases that focuses on fast and accurate memory debugging. Gained arcane knowledge of compiler internals in the process. Started off as a final-year project and was presented at two separate conferences.
  • ART.tv [Typescript, Go]: I’ve always loved films, so it only makes sense that I build a streaming platform for indie movies. Inspired by Mubi and Twitch, this is my full-stack project where I enable creators to showcase their arthouse and foreign language films to the world.
  • Cherenkov [Rust, Go]: After participating in Andreas Kling’s Browser Jam, I decided to build a minimal, full-fledged vim-based web browser. This is more of a toy project to understand browser engineering.
  • Loveless [Go, Zig]: A cool edge database that focuses on auto-scaling and performance to handle massive multiplayer environments. Started building it as part of Summer of Shipping 2024 after being inspired by Tiger Beetle and Turso.
  • Kaze Switch [Rust, OCaml]: Emulators have always been a fascination of mine and unfortunately most switch emulators struggle to run on low-end PCs. So I’m building my own lightning fast Nintendo Switch emulator with portability and networked multiplayer in mind.
  • Terra [C, Typescript]: Working on my own MMORPG as part of the Ooga Booga Game Devs community, where a group of game devs come together to ship great games from scratch, while mastering the ancient art of Caveman Programming.
  • Axogig [Python, C++, CUDA]: Building a high-performance distributed training framework to train neural nets across GPU clusters seamlessly. Started off as a project for Buildspace’s Nights & Weekends Season 5 and just kept continuing to work on it.

Future Projects include an experimental file-system to compete with ZFS & BTRFS, a GPU-aware scheduler for serverless platforms, MapReduce from scratch, context-sensitive search-engine for metadata & logs, tiny open-source machine learning compiler, library containing state-of-the-art algorithms for distributed deep learning, inference engine from scratch, building my own network stack for p2p file sharing and custom GPU orchestrator for managing H100 clusters.

Career Interests

I enjoy solving novel systems problems and prefer building reliable tools that just work and stand the test of time. This was inspired by using actually good software that never lets me down and covers all of my uses such as 7zip, Vim, Curl, i3, etc. I really appreciate and value good documentation, especially ones from SpaCy, Emacs, and Lua.

My experience has allowed me to prioritize ownership and autonomy over my work. I deeply care about my craft and tend to have healthy debates over various programming topics, especially with senior engineers. My engineering philosophy is heavily inspired by Andrew Kelley's Practical Data Oriented Design with some elements of Procedural and Functional programming sprinkled in. Here are some of my professional interests in greater detail:

✦ Supercomputers, Distributed Computing & Systems for Fun and Profit

One of my primary interests lies in designing and building novel, high-performance systems for machine learning, particularly at supercomputing scale. As a humble systems engineer, I'd kill to work on interesting projects like optimizing container runtimes, implementing distributed file systems, and building large-scale data processing applications like Apache Hadoop.
Previous experience include exploring operating system design and HPC network architectures, hacking high performance storage systems such as Weka and Ceph, designing custom load-balancing algorithm to optimize serving efficiency, breaking the CUDA compiler, and performance enhancing virtual machines.

✦ Writing Custom GPU Kernels for Machine Learning Systems

I’ve always had a deep fascination with GPUs ever since I played Crysis 3 as a kid. Now, I explore ways of writing my own high-performance GPU/CPU compute kernels for fun and that led me down a rabbit hole of optimizing CUDA kernels for TPUs, GPUs & AWS Trainium. I'd like to see how far I can push modern-day GPUs by studying the performance characteristics of various GPU architectures and optimizing compilers to leverage their hardware features like tensor cores.

✦ Managing Infrastructure and Handling SEV-1 for Breakfast

I take great pleasure in deploying on bare-metal machines and building tooling for infrastructure engineers. As a proud supporter of self-host movement, I have spent a good amount of time on private networking and cloud-native observability tools. I like being experimental with my work - e.g. integrating WebAssembly to avoid long cold starts and over-provisioning.
I'm not afraid to delve one level deeper using tools like eBPF to monitor and mitigate excessive CPU usage, instrumenting the Linux scheduler with ftrace and perfetto, analyzing request latency using sampling profilers like gprof, trying and failing to configure Kubernetes for optimal workload performance, using pprof to optimize Go code & maintain top performance across services and more spells under my sleeve.

✦ Hacking Hardware Architectures & Performance Engineering

Building robust, low-latency hardware & applications to serve billions of users has always been a career goal of mine. I would actually love to put my EE degree in use particularly on ASIC chip design and developing microprocessor architecture. I'd seize any opportunity on tearing down ML accelerators such as Google’s TPU & Groq's LPU and consumer hardware like Apple's AirTag.
I also have a good amount of battle scars in performance and backend engineering, especially in debugging kernel-level network latency spikes in containers, tuning JVM Garbage collection & developing GPU Task Schedulers, and implementing low-latency high-throughput sampling for large language models. Other strong interests include orchestration engines, block storage systems and compute services.

✦ Crafting Interactions with Love & Creative Frontend Engineering

As an interactive designer and creative front-end engineer, the web is just another canvas for me to create art and express stories. I am particularly fascinated by the challenge of delivering beautifully designed websites like those in Awwwards and Godly. I’m fortunate enough to upskill myself in various aspects of design, from typographic layouts to implementing complex web specs to get things done.
I have experience in building front-end infrastructure tooling and thrive in startupy backgrounds where ambiguity is the norm. I excel in building complex and gorgeous products, and have a high degree of empathy for user-experience, accessibility, and ownership. Taking products from 0 → 1 and designing high-quality, joy-inducing, beautiful interfaces that make a real impact are my strong suits.

Connect

“ It takes a village to raise a child ”
- African Proverb

Giving back to the community and paying kindness forward are important virtues I try to prioritize and practice. I'm always open to discussing any project or startup idea you'd want to talk about. Take a look at my socials below and feel free to get in touch for a coffee chat or collaboration. I'll be happy to chat ^-^