Ayush Verma

Researcher. Most recently I was a Computer Vision Researcher at SpyneAI and AI Engineer at IrameLabs with academic research experience during my Bachelor's.

I've worked on 3D reconstruction, efficient multimodal diffusion language models, and the infrastructure that puts ML models into production. I studied computer science at Thapar University and did research internships at Carnegie Mellon and the University of South Carolina during my undergraduate studies. I like hiking.

Ayush Verma

Work

IrameLabs, AI Engineer

2025 – 2026

Gave memory to AI Agents and made multimodals consume 90% less tokens.

SpyneAI, Computer Vision Researcher

2024 – 2025

I worked on making 3D reconstructions cleaner and faster, using splatting and diffusion refinement to remove artifacts and fill in sparse point clouds. I also built the training and inference infrastructure, improved our detection and segmentation models, and mentored six interns.

Diool, AI/ML Intern

2022

I built credit-risk BNPL models for small and mid-sized businesses in France and Africa.

Research

Carnegie Mellon University, Xu Lab

Summer 2023 · with Dr. Min Xu

I trained an object detection and tracking model for robotic arms in chemical laboratories.

University of South Carolina

Winter 2023 · with Dr. Usha Lokala and Dr. Amit Sheth

I built an LLM pipeline that helps identify aphasia, using knowledge graphs and retrieval.

Education

Thapar University

2020 – 2024

Bachelor of Engineering in Computer Science. Gold medalist in Theory of Computation.

Open source

Beta Splatting (MLX)

2026

Deformable beta splatting ported to Apple Metal MLX.

IntroToDL

2021

A PyTorch deep-learning tutorial from the ground up.

GuideToGANs

2021

A practical guide to generative adversarial networks implemented in PyTorch.

OpenCV Tutorial

2019

Human pose and video object detection from the ground up.