dheirya@deebot — ~/projects — zsh
dheirya@deebot:~/projects $ls -la --sort=interesting

stuff I actually
built & broke.

No "synergistic enterprise solutions," no fabricated 98.4% benchmarks. Just real side projects, a bit of research, and the occasional thing that only works on my machine. Honest status on every one.

4 projects ● 1 live 1 still on fire 😅 last push 3d ago
running in prod · 18h cron
notion-sync
write in Notion, ship to the blog

A TypeScript tool + GitHub Action that bidirectionally syncs my Notion database with this Jekyll blog. Runs on an 18-hour cron, preserves front-matter, and quietly deletes the posts I unpublish.

why I built it I wanted to draft in Notion (where I actually write) and stop hand-copying markdown into the repo at midnight.
TypeScript Notion API GitHub Actions Node
PRL internship · archived
Hyperbolic-Eq. Net
baking the physics into the model

Research-internship work at the Physical Research Laboratory: a neural network that folds hyperbolic equations into its architecture. Cut compute time by ~37% over the baseline while holding accuracy steady.

why I built it The real question was whether putting domain math in the architecture beats just throwing more layers at it.
Python NumPy Keras
archived 37% faster
works on my machine 😅
Agent-Eval Sandbox
personal research · ongoing

My ongoing playground for evaluating multi-agent LLM workflows — small eval harnesses with DeepEval, poking at exactly where agents fall apart, and trying to measure "did it actually work" instead of going on vibes.

why I built it My dSPACE work left me obsessed with one question: how do you actually trust an agent pipeline?
Python CrewAI DeepEval Qdrant

Currently Building!

// always shipping something
building a smarter retriever for DEE'Z-RAG — better chunking + hybrid search
shipping eval harnesses for multi-agent loops with DeepEval + CrewAI
refining notion-sync — chasing real-time instead of the 18h cron
poking at where agent pipelines actually break, and how to measure it

Contribution graph

@DheiryaBhatt