Ayush Thakur
I build and evaluate AI systems.
I am a Manager, AI Engineering at Weights & Biases by CoreWeave. I work on internal coding agents, harness benchmarks, memory, and methods for learning from previous runs.
Previously, I was a growth MLE at W&B. I built open-source integrations, production RAG systems such as Wandbot, technical courses, and practical ML writing. This work helped developers adopt W&B across training, evaluation, and LLM application workflows.
I started with autonomous robots and computer vision. Those projects taught me to work quickly, test ideas in hardware and code, and own the result. I am looking for research collaborators working on agents, memory, and learning from experience.
News
Experience in brief
Agent engineering, internal productivity, evaluation harnesses, and applied research.
Open-source integrations, LLM evaluation, Wandbot, courses, Kaggle programs, and technical education.
Technical reports on deep learning, computer vision, and practical ML engineering.
Writing and teaching
I have published 70 technical reports and 73 public notebooks across LLM systems, RAG, deep learning, computer vision, generative modeling, training infrastructure, and applied ML. I have also helped create three free technical courses and spoken at W&B events, DataHack Summit, GDG communities, and local ML workshops.
Publication
With Devjyoti Chakraborty, Snehangshu Bhattacharya, Aritra Roy Gosthipaty, and Chira Datta. We transformed heart-sound recordings into spectrograms and used convolutional neural networks to classify normal and abnormal phonocardiograms.IEEE · Google Scholar
Earlier
I studied Electronics and Communication Engineering at Netaji Subhash Engineering College, Kolkata. During college I built line followers, maze solvers, and a grid-navigation robot that avoided obstacles and retraced the shortest route.