I started with ball screws and battery packs—designing a 3-minute battery swapping system for heavy-duty electric trucks as my mechanical engineering thesis.
Then I discovered I could build systems with code, not metal.
Today, I'm a web developer at Fyle, shipping production apps by day—and a systems engineer by night, rebuilding databases and LLMs from scratch to understand how engineering principles transcend materials.
Before I wrote my first line of code, I was optimizing physical systems. This mechanical engineering thesis was my final project with atoms—and the spark that lit my journey into bits.
I realized I could iterate 100x faster with software. Design, test, deploy—all in hours instead of months. That's when the transformation began.
Live in the Wild
Today, I build software that real users depend on. Deployed. Optimized. Scalable.
Professional Timeline
Currently Building At
Sage Expense Management (formerly Fyle)
Web Developer • Marketing Automation • High-Performance Sites
Since
2024
Develop and maintain high-performance, SEO-friendly static websites using Jekyll and Astro. Partner with content marketing teams to architect and automate systems using Webflow, HubSpot, and Intercom. Focus on site speed, technical SEO, and building custom scripts to eliminate repetitive manual workflows.
By night, I rebuild the tools I use during the day. Not to compete—but to understand. Storage engines. Version control. Language models. From scratch.
Built to Understand
A learning-first storage engine built from scratch to understand database internals such as WAL, memtables, flushing, tombstones, and on-disk layouts.
A customizable PyTorch-based LLM where model size and parameters can be configured for experimentation.
A Python-based version control tool that replicates core Git functionalities like staging and snapshotting for educational purposes.
Technical Foundation
What I Share
I write in-depth technical articles documenting my journey building complex systems from first principles. My flagship work is a comprehensive 11-part series on building a Large Language Model from scratch—covering tokenization, embeddings, attention mechanisms, and transformer architecture.
A comprehensive 9-part series documenting the journey of building a Large Language Model from the ground up, covering everything from tokenization to transformer architecture.
Explores the multi-layered process of data durability, from application buffers to physical storage. A look into what really happens between clicking 'save' and data hitting the platter.