hi, i'm mihir srivastava.
Mihir Srivastava — UCSB computer engineering student, NLP researcher, and systems-focused software engineer.
i optimize for impact: real users, real traffic, real experiments.
i focus on infrastructure where correctness and latency actually matter—think auth paths, caching layers, and the observability around them.
at ucsb and beyond i've helped modernize security-critical network appliances, ship predictive caching on live logs, and explore sqlm-style rl and netlist decompilation tools.
when i'm not doing that, you'll find me trading ideas at hackathons, tearing down hardware, or trying out new agentic workflows.
recent:
- oct 2026 – presentgaucho— ucsb dining menus and macros in one view, with original nutrition labels, meal planning, a cli, and a public rest api.
- sept 2026self-play vs. text pretraining— independent follow-up to cowsik et al.'s self-play pretraining research, adding text-trained baselines and new in-context learning evaluations.
- june 2026 – sept 2026software engineering intern @ everpure— built infrastructure tooling and a custom agent harness for an internal ai triage agent, reducing false positives by 89% and increasing accuracy by 45%.
- feb 2026agent mailbox— secure postgresql-backed mcp message bus for codex and claude code agents, with direct and broadcast delivery, priority long polling, acknowledgements, and auditable events.
- jan 2026 – presentskillgen— compile llms.txt documentation into installable agent skills, with a cli, mcp server, and viewer for searching references and inspecting source coverage.
- dec 2025 – jan 2026policy radar— full-stack federal policy research agent that searches nine official government sources and streams source-backed answers with full-text and pdf retrieval.
- sept 2025 – presentundergraduate researcher @ ucsb nlp group— evaluation harness comparing single-shot, majority-vote, and persona-conditioned llm debate on aime 2025, gpqa, and hle-verified.
- aug 2025 – presentlattice— local agentic router that runs multi-step builds and tests in isolated workspaces, records structured logs, and packages only fresh deliverables.
- june 2025 – sept 2025software engineering intern @ cisco— modernized cisco asa lina ldap authentication by extracting a linux agent; moved to upstream openldap, closed 45+ cves, and made authentication 5x faster.
- sept 2024 – jun 2025decompiling higher-order functions in digital circuits— ersp research using pyrtl and egglog to recover fold abstractions from synthesized netlists and verify boolean simplifications.
- aug 2024 – sept 2024aihelp cli— python cli that converts natural-language intent into shell commands with groq-backed generation, syntax validation, and configurable execution guardrails.
- july 2024 – oct 2024software engineering intern @ spillbox— globally distributed predictive file cache built in go, trained on 100k+ access logs and achieving 95%+ prediction accuracy.
- jan 2024 – may 2024facemotion— facial-emotion classifier trained on 35k+ images with 87% accuracy, paired with gemini 1.5 pro mood recommendations; first place among 53 teams.
- jan 2024 – may 2024utility unraveled— expo/react native prototype for discovering and reporting ucsb campus utility conditions; winner of best ui/ux among 10 teams.
- jan 2024 – feb 2024wine quality classifier— compared five classifiers for red and white wine quality, reaching 84% accuracy from 11 physicochemical features and placing top five at the ucsb datathon.
- june 2022 – oct 2022software engineering intern @ openroad project | university of california, san diego— improved devops efficiency by modernizing the jenkins pipeline-interface - added features such as version tracking & performance comparison, optimized workflow & redesigned user interface for test case monitoring & development.
- may 2022 – sept 2022moore's law: its impact past, present & future— research study tracing moore’s law from 1965 through heterogeneous computing and its continuing influence on machine-learning hardware; polygence symposium finalist.