StorytellingSerious
Subrahmanyam standing on a mountain hillside with snowy peaks in the distance.

Hi, I’m Subrahmanyam.

I’m Subbu, an AI researcher and founder! Predicting numbers. Occasionally, the future. 😉

I study how AI agents make decisions as a Junior Research Fellow at Mahindra University. I also build software as co-founder and CTO of Magnocode Tech Private Limited.

NeurIPS
01

Evaluating AI agents for power-grid restoration

FLARE tests how language-model agents restore power under incomplete or conflicting observations. Physics-based checks reveal unsafe actions and failures that average performance can hide.

FLARE: Physics-Grounded Evaluation of Language-Model Agents for Feeder Restoration with Degraded Observations

NeurIPSAI Foundations for Power Grids Workshop · 2026
A branching power network isolates a faulty connection and restores downstream nodes through a pale-blue alternative path.
IJCAI–ECAI 2026
02

Evaluating EEG-to-image reconstruction

Lost in Visual Translation investigates how to evaluate whether images reconstructed from EEG preserve perceptual and semantic meaning. The work uses vision-language models and human validation to assess generated images.

Lost in Visual Translation: A VLM-Assisted Perceptual-Semantic Coherence Framework for EEG-to-Image Reconstruction

IJCAI–ECAIHBAI Workshop · 2026
EEG waveforms feed reconstructed image tiles, with separate comparison paths for visual appearance and shared meaning.
IEEE WCCI 2026 / CEC
03

Working toward safer clinical text generation

Generate, Verify and Select explores a generate–verify–select framework with evolutionary fitness shaping for clinical text generation. It is an example of my work on the reliability of generated language in an applied setting.

Generate, Verify and Select: Evolutionary Fitness Shaping for Safe Clinical Text Generation

IEEE CECIEEE WCCI · 2026
A population of abstract candidates passes through constraint checks and narrows to one pale-blue selected output.
IJSREM
04

Beginning with online proctoring

My early work examined AI-based remote examination monitoring, including intruder detection, emotion detection, and distance estimation. This area helped introduce me to research questions in computer vision and applied AI.

AI based Online Proctoring Remote Monitoring Intruder, Emotion Detection and Distance Estimation

IJSREMVol. 8 · No. 8 · 2024
A face-landmark diagram combines detection brackets, gaze rays, and distance arcs.

My code has a cosmic side.

Beyond research, I enjoy building tools that use computation and AI to make astrology easier to explore.

I’ve built Maghā Astro AI.

A comprehensive Vedic astrology platform that brings traditional astronomical calculations, advanced charting systems, and AI-assisted interpretations into one intuitive workspace.

Magha Astro AI Vedic chart workspace on iPhone
A preview of the Magha Astro AI Vedic chart workspace.

Click a tab to exploreTap a tab to explore

A pianist performing at a keyboard under blue stage lights at a university cultural festival.

I play the piano, too.

Music has grown alongside my interest in technology. I’m a Trinity College London certified pianist and have performed at university festivals and AIU competitions. I also enjoy photography and video editing.

How I got here

My interest in technology began when I was nine. Before the papers and products came experiments, a childhood science contest, a YouTube channel called STPHD, and a lot of learning by making things.

Read my story
  • A childhood photograph sitting beside a desktop computer.
  • A childhood photograph holding a phone to my ear.
  • Holding up a medal with maps on the wall behind me.
  • Presenting a headphones review for STPHD, with headphones around my neck.
  • Taking a photograph with a camera beside a lake.

What are you in the mood for?