Project 1, 2026
Starweave
A map of who influenced whom in indie music. Every link has a source.
312 artists · 1,024 links, every one with a source · Next.js 16, React 19, TypeScript, Canvas, Web Audio
Designed, built and shipped by me.
Open Starweave ↗Read the code ↗
Music apps show bands that sound alike. I wanted the ones that influenced them.
So I built Starweave.
Try it here
Starweave’s real data and code, running in this page. Every link opens its source.
Tap R.E.M. to see the proof.
Shaped by
- +7 more
Radiohead
Making music since 1985
Went on to shape
- +4 more
Photos: Radiohead by Raph_PH (CC BY 4.0)
How it works
Side BDiwakar Mishra presents
Arthouse Atlas
A film recommender for people new to arthouse cinema.
Project 2, 2026
Arthouse Atlas
A film recommender for people new to arthouse cinema. Tell it a film you loved; it picks your next.
1,120 films, ranked by 20 critics’ lists · React 19, Vite, Node and Express, MongoDB, D3
Designed, built and shipped by me.
Open Arthouse Atlas ↗Read the code ↗
For 2 in 3 people, its 10 picks included a film they went on to love.Tested on 11,090 people it never saw (66.7%, against 25.4% for the site’s first engine).
Too many films, and no idea where to start.
So I built Arthouse Atlas.
Try it here
The live site’s own recommender, running in this page on its real data.
Tap a film you love. Your next film shows below.
Your next film
Tap a film above
Your next film plays here.
Every seat is a real person who loved your film.
How it works
What’s inside, and how well it works
- Films
- 1,120 art films, 1915 to 2025
- Directors · Movements
- 238 · 27
- The canon
- Ranked by 20 lists from critics, filmmakers and festivals: 4,187 mentions in all
- Recommender
- Learns which films the same people loveEASE (Steck, 2019) · λ = 100 · 50 neighbours a film
- Trained on
- MovieLens: 32,000,204 ratings by 200,948 people
- Tested on
- 11,090 people it never saw. The test was written down before it ran, so it couldn’t be tuned to pass.
How well it works · tested on 11,090 people
2 in 366.7%
found a film they went on to love in its 10 picksthe site’s first engine 25.4% · just the most popular films 51.1%
1 in 334.8%
found the exact next film they lovedrandom picks 0.9% · first engine 7.4% · just the most popular films 20.9%
62.4%
of the films people went on to love were in its top 20 picksjust the most popular films 43.4% · “Recall@20”, the standard research measure
That's the real me, minus the pixels.
I build the logic behind the screen, not just the screen.
B.Tech CSE (AI) at MAIT, Delhi. Class of 2028. In both projects the hard part is a ranking. Starweave weighs how strong each piece of proof is. Arthouse Atlas learned from 32 million ratings what to show you next.
Works in
- C++
- TypeScript
- JavaScript
- SQL
- React
- Next.js
- Node.js
- Express
- MongoDB
- PostgreSQL
- Tailwind CSS
- Canvas
- D3
- Git