Hi, I'm Archit 👋🏽
Software Developer by craft, problem-solver and builder at heart.
Full Stack Engineer who builds end to end — from sleek interfaces to robust backends — and explores the edges of Machine Learning. Always building on the side, always sharpening my skills, and open to new challenges.
I build things — inside and outside of work. Whether it's shipping features that move businesses forward or assembling Lego on a weekend, I love bringing ideas to life. Code just happens to be my most powerful tool.
In practice, that means
Tech-agnostic by philosophy — I pick the right tool for the job, not the familiar one.
The work that isn't finished yet — pulled straight from what I've been committing lately.
Managing several git repositories at once often means juggling separate tools for issue tracking and task breakdowns — GitGrove brings that into one desktop app. Built out a branch-scoped Taskboard with full Work Item and Task CRUD, along with a Backlog list view for tracking work that hasn't hit the board yet.
I learn by building. Here are some projects that reflect how I think — check out the code and stack behind each one.
Portfolio sites reduce months of work to a static paragraph that goes stale the moment the next commit lands. Keeping a "recent work" section fresh by hand is a chore nobody keeps up with.
An agent pipeline that turns recent GitHub commit activity into portfolio-ready blurbs. It gathers activity via the GitHub API, drafts copy with Claude running headless through the Claude Code CLI, critiques the draft against a style guide, and opens a PR to this very site — nothing lands until the PR is reviewed and merged.
Résumé, pronounced by a sysadmin. The resume lives in Overleaf, but the copy deployed on the portfolio was always one version behind.
A scheduled sync agent that pulls the LaTeX source from Overleaf, uses a SHA-256 checksum gate to skip runs when nothing changed, compiles the PDF with pdflatex, and opens a PR deploying the fresh copy to the portfolio — closing superseded PRs along the way.
NYC sees thousands of vehicle collisions every year — and NYC Open Data makes that history publicly available. With over a million records on hand, it's a rich dataset for uncovering patterns in urban traffic safety.
A data engineering and analysis project focused on NYC motor vehicle collision data between summer 2019 and 2020. Built an ETL pipeline in Python to extract, clean, and load data into PostgreSQL — then queried it to uncover weekly and hourly accident trends, year-over-year comparisons, and the 100 most accident-prone consecutive days, visualized using heat maps.
Graph HyperNetworks predict neural network weights directly from computational graphs — eliminating the need for expensive optimization processes like SGD. Understanding how they degrade under limited data is key to making them practical in low-resource scenarios.
A research study analyzing how GHN-3 performance degrades across varying training data sizes using CIFAR-10 and CIFAR-100. Evaluated 80 standard PyTorch architectures and found that model architecture — not data volume — is the dominant factor in performance, with CNN-heavy residual models consistently outperforming others. Also demonstrated GHN-3's transfer learning potential across datasets.
MyAnimeList tracks anime and manga for millions of users — and with 80M+ records on Kaggle, it's a great candidate for serious data analysis.
A three-phase project comparing PostgreSQL and MongoDB on the MyAnimeList dataset — covering data modeling, query optimization, indexing, and itemset mining using the Apriori algorithm.
View on GitHubReddit is one of the most visited platforms on the web — a community-driven space for sharing, discussing, and voting on content across thousands of topics.
A full-stack Reddit clone built with React and Tailwind on the frontend, powered by Convex for real-time backend and data sync — featuring posts, votes, and community interactions.
View on GitHubIntrusion Detection Systems are a critical line of defense in network security — identifying and flagging unauthorized access before it becomes a breach.
Built a machine learning-based IDS trained on the KDD Cup dataset, leveraging Python and ML libraries to classify network traffic and detect anomalous patterns.
View on GitHubSpace Invaders — a classic arcade game that's as addictive as it gets. If you haven't played it yet, you're missing out.
A fully playable Space Invaders clone built with Python and Pygame, featuring a complete game loop, progressive level design, and a score system.
View on GitHubOpenAPI specifications power the modern web — but how composable are they really? This project attempts to replicate the findings of a 2022 IEEE paper studying data model compatibility across thousands of real-world API descriptions.
A research replication study analyzing over 71,000 OpenAPI descriptions from the APIGuru dataset. Built a pipeline in Python to clean, extract, and compare API parameters — performing endpoint matching, data-type matching, and property-name matching to validate the original paper's findings on web composability.
View on GitHubGraph databases are reshaping how we model and query highly connected data — and the query languages that power them are just as important as the databases themselves.
A literature review comparing Cypher, Gremlin, and SPARQL across syntax, semantics, optimization strategies, and use cases — examining what sets graph databases apart from relational systems and where each query language shines.
View on GitHubGot a project, an opportunity, or just want to chat? I'd love to hear from you.
Send me an Email