I work in the Pure Mathematics division at xAI (SpaceX), evaluating and extending the mathematical reasoning of frontier AI. Before that I spent twelve years running things: election campaigns, policy programmes, and academic operations at scale.
I work at the intersection of pure mathematics, frontier AI, and large-scale operations. I spent twelve years earning each part of that sentence.
Today I design research-grade evaluations of frontier-model mathematical reasoning: original problems built from contemporary mathematics literature, error analysis, and reasoning-gap completion. I grade model outputs against 50+ point rubrics, isolate the places where mathematical reasoning breaks down, and author the gold-standard solutions the model learns from. The program's hiring bar is advanced-degree mathematicians.
Before the AI frontier, I ran things. As COO of a political consultancy, I led campaign operations across three constituencies in the 2025 State Assembly Elections and won all three. At a specialist policy advisory, I was promoted from Senior Regional Manager to Assistant Director, leading nationwide policy initiatives. And at a leading EdTech platform during the pandemic's online-learning boom, I directed academic operations for 30,000+ students, helped design child-safety protocols deployed under a state-government partnership, and delivered ₹12.1 crore in cost savings.
My current focus is the emerging discipline of AI-assisted mathematical research: taking machine-suggested results to their generalized, human-verified conclusions. Publications are the goal.
Pure Mathematics division, xAI (SpaceX). Research-grade evaluation design, failure-mode analysis, and gold-standard solutions for frontier-model mathematical reasoning.
COO of a political consultancy and Assistant Director at a specialist policy advisory. Campaign operations across three constituencies, nationwide policy initiatives, teams of senior managers.
Academic operations for 30,000+ students at a leading EdTech platform, including child-safety protocols for online learning built under a state-government partnership during the pandemic. Earlier: media and communication strategy for corporate and government clients.
What it means to evaluate the mathematical reasoning of a frontier AI, why the hardest failures are the fluent ones, and what this work reveals about understanding itself.
Playing since 2017, 1,200+ games and counting. One minute on the clock is exactly enough time to be wrong with confidence.
Seven months, DEXA-verified: 93 kg at 37.4% body fat to 77.5 kg at 26.4%. Same method as everything else on this page: measure, intervene, verify.
Years of studying campaigns, from Cannae to the twentieth century. Logistics, terrain, and morale decide more than courage does.
Astronomy and quantum field theory, strictly for love. The mathematics that grades a model is the same mathematics that writes the universe.