Vishesh Yadav

Research Assistant at IISc

Indian Institute of Science | M.C.A Hons AI | Bangalore, India

Email: sciencely98@gmail.com, vishesh@nvidia.com, vishesh@corerec.tech

Summary

I am a Research Assistant at the Department of Computational and Data Sciences (CDS), Indian Institute of Science (IISc) Bangalore, where I work as a member of the AIREX Lab under the guidance of Prof. Sashikumaar Ganesan. My research lies at the intersection of computational modeling, language modeling, and machine learning, with a focus on defense-oriented applications in collaboration with the Defence Research and Development Organisation (DRDO).

Technical Expertise

Specializing in machine learning and deep learning with expertise in transformer architectures, RNNs, CNNs, and language models. Proficient in NLP, Edge AI, and model optimization.

Python C++ TensorFlow PyTorch AxLearn

Machine Learning • Deep Learning • NLP • Transformers

Contributor to Apple's AxLearn

Improved model accuracy and stability in Apple's internal AxLearn framework by contributing to quantize testing modules and fixing core SSM evaluation flows.

Thread GitHub

Jiarui Lu Ruoming Pang Mark Lee

Creator of CoreRec

Achieved 30,000+ downloads for a custom-built recommendation engine framework with support for deep learning and graph-based models.

Visit CoreRec

Part of TEGRA Team at NVIDIA

Contributed to projects at the intersection of machine learning and Autonomous Vehicles Driving in the TEGRA team.

Learn More

Creator of oioi

Ranked among Top 30 macOS utilities for developing "oioi" – a lightweight clipboard overlay used by hundreds of daily users.

Visit oioi

Developer of BHASA LLM

Designing a State Space Model–based architecture (MAMBA) for training scalable language models.

View Research

Research at IISc

Working in Prof. Sashikumar Ganesan's AIREX Lab at the intersection of finite element methods, language modeling, and computational fluid dynamics.

View Department

Deep Neural Networks Expert

Built and trained Deep Neural Networks on large-scale datasets with focus on recommendation systems, graph models, and sequence modeling.

Open Source Contributor

Active in Open Source and Research, working on projects spanning FPGA-based AI acceleration, medical NLP systems, and multilingual AI.

Current Research

IISc CDS Research

Research Assistant at IISc Bangalore, AIREX Lab under Prof. Sashikumaar Ganesan. Focus on computational modeling and language modeling for defense applications.

Computational Modeling Language Models Defense AI

DRDO Collaboration

Defense-oriented applications in collaboration with DRDO. Machine learning research for national security applications including threat detection and autonomous systems.

National Security Threat Detection Autonomous Systems

Research Focus

BHASA LLM

Developing BHASA LLM using MAMBA architecture. Research on state space models, Bayesian approaches, and efficient AI architectures with sustainable applications.

MAMBA Architecture State Space Models Bayesian AI

Edge AI Research

Optimizing AI models for edge devices with focus on real-time processing, model compression, and hardware acceleration for defense applications.

Edge Computing Model Compression Real-time AI

Core Frameworks

CoreRec Framework

A custom framework similar to PyTorch/TensorFlow for recommender systems. Provides modules to build scalable recommendation systems with DNG scoring. CoreRec has over 22,000 installations.

Python Recommendation Systems DNG Scoring 22K+ Installs

Custom ML Pipeline

Built end-to-end machine learning pipelines for defense applications, including data preprocessing, model training, and deployment optimization.

ML Pipeline Data Preprocessing Model Deployment

Technical Stack

Deep Learning Frameworks

Expertise in PyTorch, TensorFlow, and custom implementations. Specialized in transformer architectures, RNNs, CNNs, and state space models.

PyTorch TensorFlow Transformers Custom Implementation

Optimization Techniques

Advanced model optimization including quantization, pruning, and knowledge distillation for efficient deployment on edge devices.

Model Quantization Pruning Knowledge Distillation

AI Applications

SLYRIC Project

Created SLYRIC (Sign Language Yielding Realtime Intelligent Classifier) as an independent project for real-time sign language classification, optimized for edge devices.

Computer Vision Real-time Processing Accessibility Edge AI

Autonomous Vehicle AI

Worked on DriveNet for autonomous vehicles at NVIDIA, focusing on data preprocessing and model optimization for real-time decision making.

Autonomous Vehicles Computer Vision Real-time AI

Specialized Applications

Defense AI Systems

Developing AI systems for defense applications including threat detection, autonomous navigation, and decision support systems.

Threat Detection Autonomous Navigation Decision Support

NLP Applications

Natural language processing applications for defense communication, document analysis, and multilingual text processing.

NLP Document Analysis Multilingual

Apple Collaboration

Working with Farzad Abdolhosseini on Intra Language Interfaces. Successfully merged contributions to Apple's AXLearn, optimizing model architecture and testing.

Apple AXLearn Model Optimization Open Source

NVIDIA TEGRA Team

Worked at NVIDIA in the TEGRA Team preprocessing data to train DriveNet for autonomous vehicle driving.

NVIDIA TEGRA DriveNet Data Preprocessing

IISc AIREX Lab

Research collaboration with Indian Institute of Science, AIREX Lab under Prof. Sashikumaar Ganesan, focusing on computational modeling and AI research.

IISc Bangalore AIREX Lab Academic Research

DRDO

Working with Defence Research and Development Organisation for defense-oriented AI applications and national security research.

DRDO Defense Research National Security