Chanda Charan Reddy.
I build intelligent systems.
I’m Charan — an AI & automation engineer in Bangalore. I ship production LLM systems: a Springer-published model that reads chest X-rays well enough for a radiologist to take seriously, document pipelines that quietly run themselves, and the occasional thing like this one — which exists mostly because I couldn’t stop thinking about it.
Before any of that, I wrote real-time control code for jet engines at DRDO, where a millisecond of lag isn’t a bug — it’s a flameout. That job rewired how I think about latency and failure. You’ll find its fingerprints all over Synapse: the obsession with what happens when the network drops, the refusal to block the user on a round-trip, the quiet paranoia about state.
This is one project. There are 18 more (and a few jet engines) over at charanreddy.dev.
Five things Synapse taught me.
- 01
CRDTs look like magic in a notebook and fall apart in production.
The day I stopped my own updates from echoing back through the relay was the day “multiplayer” stopped meaning “infinite loop.” Convergence isn’t the hard part — origin discipline is.
- 02
“Offline-first” is a stance, not a feature you bolt on.
Once the local Yjs doc is the source of truth and the network is just an optimization, half the hard questions answer themselves. Reads and writes stop waiting for permission.
- 03
End-to-end encryption is mostly about what you refuse to send.
The key never leaves the URL fragment; the server routes by a hash it can’t reverse. Most of the work was deleting trust, not adding cryptography.
- 04
Good architecture pays you back in features you didn’t plan for.
Time-travel was almost free — because the CRDT update log already is the history. I just had to replay it. Choose the data model well and the demo writes itself.
- 05
Building with no users is its own kind of honest.
No dashboard to hide behind. Either the offline edits merge cleanly in front of you, or they don’t. That’s a good way to keep yourself from lying in a README.
Have an idea? Let's talk.
Want to build something — or break something interesting? I'm always up for a good problem. Book a call, send a note, or just say hi.