Skip to content
Siddesh Sambasivam Suseela

Hi, I'm Sid 👋

Engineering lead at Hypotenuse AI, building LLM infrastructure for ecommerce product data at scale.

Work

Apr 2026 - Present
Lead Software Engineer · Hypotenuse AI
  • Lead the engineering team behind a SOC 2 platform processing millions of SKUs a day.
  • Cut categorization cost 80% at 95%+ accuracy by training an embedder and reranker.
  • Split read and write stores (CQRS), cutting worst-case reads from 50s to under 1s.
  • Kept the read store within 1s of writes at p99 using a transactional outbox.
  • Cut per-product cost 43% after tracing an 11x spike, using model routing and batching.
  • Gate 100% of prompt and model changes on evals with human-calibrated LLM judges.
  • Built agent profiles that rerun evals on every change to catch model regressions.
  • Feed customer ratings and edits back as evals, with an agent drafting retraining plans.
  • Built a sandboxed multi-agent harness that beats frontier lab harnesses on our evals.
  • Shipped to enterprise retailers with catalogs of up to 15M+ SKUs, discovery to rollout.
  • Took onboarding to zero custom code with an agent that builds each customer's pipeline.
  • Own hiring, mentoring and design review, plus the multi-region AWS expansion.
See the full work history

Selected publications

Google Scholar
ICARCV 2022 Dec 2022
Read
A Comparative Study on Machine Learning Algorithms for Knowledge Discovery
A survey and head-to-head benchmark of symbolic regression methods — Genetic Programming, Deep Symbolic Regression, AI-Feynman, and NeSymRes — on the Feynman-03 and Nguyen-12 datasets. We highlight the strengths, failure modes, and open problems for each family of methods.
S. S. Suseela, Y. Feng, K. Mao
Nanyang Technological University 2022
Read
Scientific Machine Learning for Knowledge Discovery
Bachelor's thesis at Nanyang Technological University. Investigates how machine learning methods can recover interpretable, closed-form models from data — bridging the gap between black-box prediction and scientific insight.
S. S. Suseela