Sanjay Manivasagam Applied AI / ML Engineer
I build AI agents, production LLM workflows, and the infrastructure to evaluate them.
UC Davis · B.S. Data Science, 2026 · San Francisco Open to full-time roles and relevant internships
LLM reporting for 200+ monthly reports at DataCorp Traffic.
Agent infrastructureActions → evidenceAudit records, verification, and human oversight in ClearAgent.
Evaluation rigorTest the testBuilt Groundtruth to measure detector errors and challenge its own conclusions.
I built audit and oversight mechanisms that make agent decisions inspectable. Explore how a record is checked and when a decision needs human review.
My contributionBuilt the audit and oversight infrastructure; scoped which agent decisions need evidence and human review.
Public TypeScript implementation · Separate private LangGraph work
I built a retrieval and evaluation harness that compares matched conversations under shared judge context. The experimental design is inspectable; its treatment effect is withheld.
My contributionBuilt the two-channel retrieval harness and owned reproducibility for a four-person team.
4 experiment legs · 2 direction-matched judge batches
I built a deterministic red-teaming harness, measured its detector errors, and traced an apparent safety result to the harness itself. Changing the adapter changed the conclusion.
My contributionBuilt the harness and detectors, labeled the evaluation corpus, and investigated the state-handling confound.
68 hand-labeled traces · Mocked, deterministic tools
None of the tested interventions beat the strongest zero-shot baseline. Explore the prompting and fine-tuning conditions, and the separate transfer experiment.
My contributionBenchmarked prompting and fine-tuning, tested cross-dataset transfer, and analyzed prediction failures.
11 configurations · 2 seeds · One model family
We built a database engine around versioned columnar storage. Follow a record through an update, then explore what happens when transactions compete for it.
My contributionBuilt the HTAP engine with a five-person team, covering versioned storage, buffering, merge, and concurrent transactions.
Record-level locks · No independent throughput benchmark
A recognition score is only part of the story. Compare correct identifications with unknown people mistakenly accepted at two reported operating points.
My contributionOwned testing and comparison design, predictions, and metric analysis, rather than every model implementation.
Two reported operating points · Report and presentation available
I build AI agents, production LLM workflows, and the infrastructure to evaluate them. At ClearAgent, I developed audit and oversight mechanisms and a separate private LangGraph orchestration service. Previously, I shipped a production reporting pipeline at DataCorp Traffic supporting 200+ monthly reports and reducing delivery from days to hours. My work connects applied AI with backend engineering, data systems, and careful measurement.
AI Engineer · San Francisco, CA (Remote)
Built agent audit and oversight infrastructure, plus a separate private LangGraph orchestration service. Independent engineering project.
AI/ML Engineering Intern · Remote, UK
Built a production LLM reporting pipeline supporting 200+ monthly reports, reducing delivery from days to hours. Owned validation through authenticated APIs as one of two developers.
Software Engineering Intern, Agents & Frontend · San Francisco, CA
Extended an AI coding agent across multi-file repositories and shipped a React/TypeScript review dashboard, cutting clicks-to-accept by 50%.
JUL 2, 2026 · 2 MIN READ
Eleven conditions of chain-of-thought prompting and fine-tuning on table reasoning. The baseline won. Here is why that is worth reporting.
JUN 10, 2026 · 3 MIN READ
Turning broad oversight obligations into inspectable records, integrity checks, and review triggers.
