Ritesh Sharma.
senior software engineer · tech lead · java distributed systems · india
Hi — I'm Ritesh. For the past eight years I've been the engineer teams call when a trading platform can't go down. I lead hands-on — equal parts whiteboard, JVM internals, and mentoring — and my favorite wins are the unglamorous ones: the restart nobody worries about anymore, the dashboard that stays green.
Lately I've fallen for a new version of an old problem: making agentic AI dependable — building predictable systems out of parts that aren't. The old-school disciplines turn out to carry over: distrust every input, verify every output, design for failure before success. Different tools, same craft.
Work & Writing
-
Cutting a Tier-1 trading platform's recovery from 90+ minutes to 7
Parallel Cassandra token-range loading, Ignite cache topology, and staged cluster formation — a 13× faster restart path for a multi-region platform.
4 min read -
A hash collision, a wrong diff, and my first open-source PR
JaVers was silently dropping changes from diffs because two global IDs can share a Java hashCode. Debugging it in production and fixing it upstream (merged, released in 7.8.6).
3 min read
About
I like problems where correctness and uptime actually matter — the kind where you read database internals on a Tuesday because production needs you to. I stay hands-on while leading: architecture reviews, mentoring, Cassandra and JVM tuning, and Sev-1 on-call for systems clearing 10M+ trades a day.
What I've learned from the AI work: the hard part isn't the model, it's turning a messy human workflow into an engineering problem worth solving. I want more of that.