Lalith Suresh
banner
lalithsuresh.bsky.social
Lalith Suresh
@lalithsuresh.bsky.social
CEO @ feldera.com, the incremental compute engine for AI, ML and data teams.

Formerly a systems researcher in distributed systems, databases, cloud, OS, PL, and networking. Sci-fi and gaming nerd.

lalith.in/research
Trying to pick a game to play and this is one of claude's suggestions.
October 26, 2025 at 7:20 PM
Claude, relax.
October 11, 2025 at 12:33 AM
nice form validation
June 29, 2025 at 6:35 PM
Staring in wonder watching Mihai pull this off in under a week. 🤯

github.com/feldera/feld...
June 28, 2025 at 1:05 PM
It really goes from 0 to thunderstorm here in Atlanta.
June 27, 2025 at 11:24 PM
Our Rust compilation shenanigans have hit HN
April 17, 2025 at 2:36 PM
#1 on HN now
March 20, 2025 at 5:34 AM
Spring cleaning my apartment and thought I'd share this gem. It's the t-shirt we designed for my Masters program in distributed systems, many moons ago. #emdc
March 20, 2025 at 3:10 AM
Spring cleaning some docs 🧹🧹🧹🧹
March 19, 2025 at 5:08 AM
One of my cats, who's always otherwise asleep during the day, decided to make a guest appearance mid podcast a few times.
March 18, 2025 at 4:31 PM
One example from the paper:
March 1, 2025 at 12:07 AM
We've got some artists here: xls.feldera.io

🌞
January 16, 2025 at 7:16 PM
It's ain't much but it's honest work.

github.com/feldera/feld...
January 16, 2025 at 7:06 PM
I’ve worked on performance engineering for ~13 years. Queuing theory has been my go-to framework the entire time -- thinking in terms of arrivals, departures, queues, servers, service rates, latencies, waiting times, and priorities. It’s simple to apply and doesn’t require diving into the math.
January 3, 2025 at 12:20 AM
This is how I mapped the literature in this space:
December 23, 2024 at 7:02 PM
Using Konsole within GNOME!?
December 19, 2024 at 7:52 PM
If you're into board games, you should definitely try Wingspan!
November 27, 2024 at 2:29 PM
Me waiting for Java's Project Valhalla to ship
November 25, 2024 at 7:28 PM
This is control plane code so it is hardly performance sensitive. It's well worth it in my opinion to do such a refactoring. Anvil did require controllers to be written in that style and there was practically no performance difference (see 7.3 in the paper).
November 19, 2024 at 4:08 AM
You already construct one ring deterministically right? Do that for K rings out of the same memberlist (e.g., with K different fixed seeds). Each member monitors one member next to it per ring. This process produces random graphs that is known to yield expanders with high-probability.
November 12, 2024 at 11:20 PM