Poorani Muthusamy
Seattle-based engineer with 15+ years building mission-critical systems — from ISRO satellite and lunar-rover operations to enterprise platforms serving ~10,000 Starbucks stores. Independent researcher in causal inference for systems reliability.
profile.jpg in the same folder as this page.
governed
A small, explicit framework for goal-directed LLM agents. Every step is a contract the model must satisfy before it can proceed: plan before acting, cite evidence before planning again. The result is an agent whose every action is traceable to a stated reason.
View on GitHub →Research Interests
Learning Causal Structure for Multimodal Anomaly Detection in Microservices. My current reading and early work centers on:
I'm looking to collaborate with people working on any of these — on the methods side (causal discovery, RCA) or the systems side (real telemetry, incident data). Open to co-authorship, reproduction studies, and benchmark work.
About
My career began at the Indian Space Research Organisation (ISRO), where I spent over a decade owning ground-segment software for satellite operations — work where a single defect could jeopardize an operationally irreversible mission. I led teams delivering ground software for the Chandrayaan lunar-rover program, architected autonomous satellite scheduling platforms, and led an ML-driven anomaly-detection research effort on real-time spacecraft telemetry. My work on semi-automated path planning for the Chandrayaan-2 rover had an abstract accepted at the 43rd COSPAR Scientific Assembly (2021).
Today, at Starbucks, I build the enterprise Workforce Management platform on Java, Spring Boot, Azure, and Databricks — powering scheduling, forecasting, and compliance for roughly 10,000 stores. Along the way I've hardened payment systems at Costco Travel, advised an early-stage fintech through its first funding round, and championed responsible AI-assisted development across my teams.
Experience
Recognition & Impact
Abstract accepted — COSPAR 2021
Abstract on Chandrayaan-2 rover path planning accepted for the 43rd COSPAR Scientific Assembly, an international forum of the Committee on Space Research.
National space program
Mission-critical ground software for India's lunar-rover and satellite operations across 11+ years at ISRO.
Enterprise scale
Systems serving ~10,000 Starbucks stores and tens of thousands of employees across North America.
Payments & compliance
PCI-DSS-compliant payment microservices at Costco Travel; advised a fintech startup to its first funding round.
Responsible AI
Authored team-wide standards governing AI-assisted development and led an internal AI assistant prototype.
Leadership
Led research and engineering teams across aerospace and retail; mentors engineers through design review.
Recommendations
Poorani is one of the most outstanding engineers I've worked with. She takes the time to understand the business problem, customer impact, and technical landscape before building solutions, ensuring lasting results rather than quick fixes. Her ownership, technical excellence, and focus on outcomes make her a trusted leader and collaborator.
I had the pleasure of working with Poorani at Starbucks, where she led the project I was part of. What stood out was how she balanced trust with support — she stayed close to the work and always knew where things stood, but never micromanaged. She gave me the space to think through problems on my own, and whenever I hit a wall, she'd step in with a precise, well-aimed solution that unblocked me right away. A sharp engineer and genuinely enjoyable to work alongside. Any team would be lucky to have her.
Writing
Our Children Are Learning With AI. We Should Decide How — On Purpose.
A parent-and-engineer's case for a deliberate, tiered, purpose-first framework for AI in schools.
Zero-Downtime Database Migrations
The expand–migrate–contract pattern for evolving a live database — renames, constraints, indexes — with no downtime.
The Day I Had My I, Robot Moment
An honest, funny field report on living and working alongside an AI assistant.