Hello Everyone!

I am Pratik Pranav, currently working as an AI Engineer at ThirdAI Corp. At ThirdAI, our focus is on training machine learning models on CPUs using dynamic sparsity. My role involves the intersection of machine learning and systems, emphasizing the practical and scalable implementation of machine learning technologies.

---

My Work

Distributed Training Framework for BoLT: In my first six months at ThirdAI, I worked on implementing the Distributed Training Framework for our internal machine learning library, Bolt. My work also included to incorporate distributed training in other ThirdAI offerings, such as the Universal Deep Transformer (UDT) and ThirdAI’s NeuralDB. Here is a joint article with Anyscale detailing our work: Anyscale Blog Post.

Bolt2.5B: Lately, I have worked on developing ThirdAI’s LLM, Bolt2.5B which includes the architecture design, training, and deployment. This project was challenging due to its emphasis on CPU-based training, necessitating innovative system solutions. For more details, see the ThirdAI Medium article: ThirdAI Medium Post.

---

Papers

BOLT: An Automated Deep Learning Framework for Training and Deploying Large-Scale Search and Recommendation Models on Commodity CPU Hardware
Authors: Nicholas Meisburger, Vihan Lakshman, Benito Geordie, Joshua Engels, David Torres Ramos, Pratik Pranav, Benjamin Coleman, Benjamin Meisburger, Shubh Gupta, Yashwanth Adunukota, Siddharth Jain, Tharun Medini, Anshumali Shrivastava
Accepted at ACM International Conference on Information and Knowledge Management 2023

From Research to Production: Towards Scalable and Sustainable Neural Recommendation Models on Commodity CPU Hardware
Authors: Anshumali Shrivastava, Vihan Lakshman, Tharun Medini, Nicholas Meisburger, Joshua Engels, David Torres Ramos, Benito Geordie, Pratik Pranav, Shubh Gupta, Yashwanth Adunukota, Siddharth Jain
Accepted at ACM Conference on Recommender Systems 2023

---

Check out this website’s source here!
Hosted on GitHub Pages with Jekyll.
Homepage theme on page-themes.