Nitesh Methani
I'm Nitesh Methani. I build durable, reliable AI systems for production. My work spans agent runtimes, evaluation, orchestration, observability, and inference—the infrastructure that turns frontier models into enterprise-grade products.
Previously, I built long-running AI agents at Sarvam, worked on address intelligence and geospatial ML at Amazon, and began in research at IIT Madras, advised by Mitesh Khapra and Pratyush Kumar.
For enterprise AI systems and technical collaborations, get in touch.
Separating signal from noise in financial news
A framework for separating consequential information from market chatter, and for knowing when a headline should actually change your view.
The pessimist VC's 15-point kill list
A red-team checklist for startups: the questions that expose fragile ideas before a compelling story takes hold.
Tayari
A GATE CS preparation app built around focused practice and recall, now used by more than 10,000 students.
Learning geospatially aware place embeddings via weak-supervision
A SIGSPATIAL paper on combining delivery history, address text, and geography to resolve ambiguous real-world locations.
A Systematic Evaluation of Object Detection Networks for Scientific Plots
An AAAI study of how object detectors read scientific plots, and why charts demand more precise localization than natural images.
PlotQA: Reasoning over scientific plots
A WACV benchmark that made visual reasoning over scientific plots a rigorous, measurable problem for machine learning systems.