Research benchmark
NepSyc
Six kinds of sycophancy, scored separately across English, Nepali and Romanized Nepali.
705test items
3language splits
6behaviours
- Python
- Streamlit
- Groq
- LLM-as-judge
Projects
Everything I’ve built
Four case studies, plus smaller experiments. Filter by kind.
Full-stack product
ClipGen
Authentication, rate limits, Stripe payments and AI video generation in one Next.js app.
- Next.js
- tRPC
- Prisma
- Neon
- Redis
Real-time app · live
Chatly
WebSockets, Google sign-in, image sharing, unread counts and live presence. Deployed.
- MongoDB
- Express
- React
- Node
- WebSockets
Systems · networking
LazyFlare
A and CNAME records, zone files in a folder, well-formed responses.
- Rust
- DNS
- Zone files
- A / CNAME
Also
Smaller experiments
NepaBias
Team benchmark of gender bias in LLMs across Nepali recruitment, scholarship and recommendation-letter text. Gemini 2.0 Flash, XLM-RoBERTa, multilingual MiniLM.
On GitHub ↗NEPSE gain tracker
Next.js, Prisma and PostgreSQL app for profit and loss on Nepali stock holdings, with local broker-commission and tax rules.
On GitHub ↗