Varad Gunjal
Applied Scientist at Amazon in the AGI org working on research in pretraining multimodal foundation models. I am part of the core team that developed the Nova model family.
I graduated with a Masters in Electrical & Computer Engineering from Carnegie Mellon University where I spent time at the Parallel Data Lab.
I completed my undergraduate degree in Electrical & Electronics Engineering from BITS Pilani, during which I was fortunate to have the opportunity to do a Bachelor Research Thesis at Institute of Neuroinformatics, ETH Zurich.
Where I've Worked
Applied Scientist
Architect & train the vision encoder for the Nova family of multimodal foundation models, and research data-centric techniques for improving their performance and efficiency.
Robotics Software Engineer
Building image processing applications - orthomap generation and vision-guided autolanding - for the production drone platform, Rapyuta Connect.
Co-Founder
Co-founded and led engineering for an image-based search and retrieval engine for e-commerce.
R&D Intern
Prototyped inverse kinematics solver in C++ using Jacobian pseudoinverses for the internal animation tool - Premo.
View my full CV here.
Academic Background
Carnegie Mellon University
University Name
Technical Knowledge
Languages & Frameworks
- Python, C/C++
- PyTorch, JAX
- numpy, scipy, pandas
- PySpark, Ray
Hardware
- GPU, CUDA, Triton
Backend & Infra
- PostgreSQL, Redis
- Docker, Kubernetes
- AWS