Siva Rajananda

Senior Data and Applied Scientist at Pattern, New York

I build production ML and LLM systems, with a researcher's habit of checking whether a system knows when it is right.

• ~20Mpredictions a day at peak, on the largest pipeline
• ~600ksimulations per run in a decision engine I designed
• Live pilotfor an LLM customer-service agent I took over and reworked
• 8 paperspeer-reviewed, on confidence and perception

Work

• Conversion-rate prediction for ad bidding

  The models behind live ad bids on three marketplaces, with data checks that stop a bad run before it reaches the bids.

• An LLM customer-service agent

  Reworked a pre-production agent and took it to a live pilot, testing every change against real customer questions.

• A decision engine for excess inventory

  A simulation in production since 2024. In 2026 I cut a full run to about 40 minutes.

• An MCP server for AI agents

  Lets internal AI agents answer inventory questions, and never sound more certain than the data.

Research

Before industry I spent about three years in cognitive science labs asking one question: does a person know when they are right? I also built the first open browser version of a standard vision experiment, used by labs around the world since 2017. Try the experiment and read the research.

Education

• 
  Harvard UniversityM.S., Data Science
  2020 to 2021
• 
  UCLAB.S., Cognitive Science, computing specialization
  2013 to 2017

Experience

• PatternSenior Data and Applied Scientist. Lehi, Hong Kong and New York.
  Since 2022
• ASMLData science intern
  2021
• UCLA and UC RiversideResearch associate, lab manager and programmer
  2017 to 2020

• Resume (PDF)
• More about me
• LinkedIn