Yashaswi Mishra
I spend most of my time in Go and Rust, building backends that don't fall over and systems that scale when they need to.
I'm drawn to databases, distributed systems, and the kind of problems where you have to think about what happens when things go wrong.
When I figure something out, I write about it. When an idea won't leave me alone, I turn it into an open source project. I've been fortunate to work with people who take craft seriously, and if any of this resonates, you can .
Now
A snapshot of what I'm focused on right now. Reading, thinking, building.
Database Internals
Alex Petrov
Efficient and Robust Approximate Nearest Neighbor Search Using HNSW
Yu. A. Malkov, D. A. Yashunin
The Adaptive Radix Tree (ARTful Indexing)
Leis, Kemper, Neumann
HNSW is beautiful — a multi-layer probabilistic skip list of nearest-neighbor graphs. The whole thing rests on one idea: long-range links at the top, fine-grained refinement below.
mmap is the storage engineer's nail gun. Powerful, fast, and easy to put through your own foot if you don't respect fsync ordering.
Things I've built that I'm proud of.
kova
A distributed vector database, built from scratch in Rust.
SIGKILL-tested across 55,697 acked inserts · ~16× faster than flat at 100k vectors
pastel
skribbl, but the game is the call.
One tokio actor per room · hand-rolled postcard codec · 35 ms p50 on Cloud Run
juzfs
Google's File System paper, in Rust.
GFS architecture from scratch in Rust · 5-day sprint · raw TCP framing protocol
plethora
Amazon's Dynamo paper, in Go.
Strategy-3 consistent hashing · vector clocks · sloppy quorum + hinted handoff · Merkle anti-entropy
How Databases Compact Without Blocking Reads (LSM Engine Design)
How LSM trees keep reads available while compaction runs in the background.
CRDTs Explained: Building a Collaborative Whiteboard from Scratch
Conflict-free replicated data types, built up from first principles into a real collaborative whiteboard.