Yashashwi
Singhania
I build systems and overthink algorithms. Dual degree at IIT (BHU), mostly writing reinforcement learning, systems code, and native macOS apps people actually keep installed.
- based in
- Varanasi, India
- studying
- Biochem. Eng.
- graduating
- 2028
- shipping
- macOS apps
1720
Codeforces
hackerman15
8.74
CGPA
IIT (BHU)
13th
AMS Derive 2026
All India Rank
3000+
Participants
Byte the Bits
Selected
work
11 repositories
most of them still running
Fadeo
A macOS menu bar app that watches your workflow — the app in front, the desktop you're on, whether you're in a meeting — and plays, fades, or switches audio to match. Every rule is yours to define, down to the second.
- Swift
- SwiftUI
- macOS
- Audio
Arras
A macOS menu bar app that places photos on your desktop as borderless, always-on-desktop overlays — perfectly matching each image's native aspect ratio. No cropping, no black bars.
- Swift
- SwiftUI
- AppKit
- Core Animation
CodeForge
A competitive programming platform with Codeforces integration. Host custom contests, track performance analytics, get personalized problem recommendations, and compete on ICPC-style leaderboards.
- React
- Node.js
- MongoDB
- Codeforces API
FFA-Pacman
Custom multi-agent Pacman environment built from scratch using PettingZoo. Agents compete to collect energy, eliminate each other, and survive — trained with DQN via Stable Baselines3.
- Python
- Deep RL
- PettingZoo
- SB3
A small studio
PureMac
Two native macOS apps, both open source, both free to run. Arras puts photos on your desktop at their real proportions. Fadeo changes your audio to match what you are doing.
Where I have been
June 2025 — July 2025
Remote
Machine Learning Intern
Fxis.ai
- Architected a multimodal RAG pipeline utilizing Weaviate to parse complex enterprise documents.
- Engineered advanced embedding models to seamlessly extract structured text and tabular data.
- Automated analytical report generation from UK corporate tax filings to accelerate compliance.
- RAG
- Weaviate
- NLP
- Embeddings
Aug 2024 — Nov 2024
Varanasi, India
Research Project
IIT (BHU) Varanasi
- Explored theoretical aspects of two-agent reinforcement games without shared states.
- Developed value/policy iteration algorithms for non-shared state systems, proving convergence to Nash equilibrium.
- Simulated adversarial dynamics in competitive settings and analyzed resulting strategic behaviors.
- Reinforcement Learning
- Game Theory
- Nash Equilibrium
Languages
- Python
- C++
- JavaScript
- TypeScript
- LaTeX
Web & Frameworks
- Next.js
- React
- Flask
- Tailwind CSS
- SwiftUI
- AppKit
Tools & Platforms
- Docker
- Git
- GitHub Actions
- Vercel
- Google Colab
- SSH
- Bash
Operating Systems
- Linux
- macOS
Receipts
Programming & Hackathons
- 1720Codeforces Ratinghackerman15
- 1stPrize — DevBitsUdyam, IIT BHU 2025
- 13thPOSTERIOR — AMS Derive 2026All India Rank · Top 50 National
- 39thPRIOR — AMS Derive 2026All India Rank · Top 50 National
National Examinations
- 1105thKVPY-SA 2021–22All India Rank · Kishore Vaigyanik Protsahan Yojana (Stream SA)
- 20thNTSE — Uttar PradeshState Rank · National Talent Search Exam
- QualifiedIOQM 2021Indian Olympiad Qualifier in Mathematics
Also built