Srikar Chittemsetty
I study computer science and philosophy at the University of Chicago, and I'm a founding software engineer intern on the infrastructure & platform team at ForkLaunch. I build systems that stay correct when things fail: exactly-once execution, ingest that survives overload, and benchmarks that check whether an AI agent actually finished.
Activity
All updates →We won, what now? — Will Wilson, Bug Bash 2026
Intro to Large Language Models — Andrej Karpathy
Rebuilt Gray's process pair and measured the checkpoint interval across 420 failovers — process-pair, GitHub
Experience
ForkLaunchFounding Software Engineer Intern, Infrastructure & PlatformOwned observability and alerting in production. Merged 51 pull requests and 43,000 lines in the first three months, from the alerting pipeline and a Rust CLI to the monitoring dashboard.Jun 2026 – PresentInternship · San Francisco, CA
University of Chicago, Khomtchouk & Grossman LabsResearch Software EngineerAutomated the lab's batch data pipeline in R as jobs on a shared Slurm cluster, caught silent data corruption from mismatched field IDs between upstream sources, and ported the pipeline to NIH's All of Us biobank, validating 9 datasets across 500k+ records.Oct 2025 – Jan 2026Part-time · Chicago, IL / RemoteProjects
All projects →Selfbench
2026
A benchmark that turns your own commits into tasks for coding agents and grades them with the tests the real commit added, which the agent never sees. Run on 15 fixes merged into Warp's open-source repo, Warp's Oz agent said done in 30 of 30 runs and failed the hidden tests in 10.
Python / AI Agents / Evals / Hidden Tests
Process Pair
2026
Jim Gray's 1985 process pair rebuilt on one laptop and killed 420 times to measure the checkpoint tradeoff he described. Checkpointing every request costs 25% of throughput, lost work is N/2, and the 1974 formula for the best interval lands 20x too high because it assumes lost work gets recomputed.
Python / TCP / Fault Tolerance / Distributed Systems
Hapax
2026
A crash-durable task store for AI agents that guarantees a side effect happens exactly once, even under kill -9 at any point. 500 randomized crashes produced 0 double charges; a naive implementation under the identical harness produced 121.
Python / Go / PostgreSQL / Distributed Systems
Aporia
2026
Semantic search over primary-source philosophy that classifies whether a passage argues for or against a claim, not merely whether it is about the topic. On a held-out set written after the fix, 20 of 21 queries surfaced the philosopher who actually holds the position.
Python / FastAPI / Next.js / TypeScript
Peras
2026
A metrics ingest pipeline that survives the label explosion that OOM-kills a naive one, proven by exit code. Same 20M-sample workload in the same 256 MB container: the naive ingester dies at 8M samples; Peras completes all 20M at 1.67M samples/s with ~3 MB of heap.
Go / gRPC / OpenTelemetry / Kubernetes
Open Source
Merged pull requests to Kubernetes, OpenTelemetry, and Microsoft projects.
Render negative Pod priorities as valid label values
kubernetes-sigs/jobset
Bound decompressed size of gzip request bodies
open-telemetry/opentelemetry-collector-contrib
Map the Rate metric intake type in the Datadog exporter
open-telemetry/opentelemetry-collector-contrib
Clarify processor ordering relative to batching
open-telemetry/opentelemetry-collector
Add TypeScript SDK quickstart example
microsoft/agent-governance-toolkit