Now
What I'm working on now.
Updated
Finishing my degree
I graduate from Northeastern with an M.S. in Data Science in December 2026, and I'm looking for forward-deployed engineer, data scientist and software engineer roles. The resume has the details.
Teaching
I'm a teaching assistant for DS 3500, Advanced Programming with Data, with Professor Deahan Yu. A lot of it so far is helping students through Git for the first time, merge conflicts included. Before this, I spent January to July 2026 as a data scientist at Launch Plan; the case study covers what I built there.
Research
Getting “Predictive Variance Decides”, the corrected AURORA paper with Alekhya Bulusu, compiled and onto arXiv. The case study explains how the original 41% gain turned out to be a bug, and what held up instead.
Building
A series of small studies on how LLM systems fail, Measuring LLM Systems: so far one poisoned passage in a RAG corpus, stale agent memory, 22,743 A/B tests re-analysed, how much harmless rewording changes answers and what sways an LLM judge. Next: long-context recall, then tool-call reliability. Alongside it, whygit is on PyPI and the MCP registry, and GitGotchi 0.4.0 is out.
Writing
Case studies of things I've built, including what went wrong, on the Writing page, and posts on my LinkedIn newsletter, Vectorized Thoughts.
This is a now page.
Get in touch
Questions about my work, a role, or a project idea. Send a message here or email me directly.
Srikarmks@gmail.com