Professional Experience

Imaging Lyceum Lab, Arizona State University

Research Assistant

Tempe, AZ

January 2024 - Present

  • • Developing state-of-the-art 3D reconstruction and rendering techniques leveraging Gaussian methods, deep learning, and implicit neural representations.
  • • Designing multi-modal sensor fusion techniques for improved 3D scene understanding, integrating Time-of-Flight (ToF), Acoustic, Electromagnetic, and Optical data.
  • • Optimizing algorithms for high-performance computing (HPC) environments, utilizing CUDA, PyTorch, and parallel processing for large-scale 3D reconstruction.
  • • Collaborating with cross-disciplinary teams in computer vision, acoustics, and computational imaging to advance next-generation imaging technologies.

Dartmouth University

Visiting Doctorate Student

Hanover, NH

May 2024 - July 2024

  • • Developed and prototyped advanced Acoustic-Non-Line-of-Sight (NLOS) imaging hardware and algorithms, enhancing visibility through occlusions using deep learning and signal processing.
  • • Investigated multi-modal imaging approaches, particularly Acousto-optic fusion, integrating acoustic and optical data to refine depth estimation and scene reconstruction.
  • • Conducted extensive experimentation on wave propagation models, improving acoustic-based scene inference in complex environments.

RAIN AI

AI Hardware and Systems Intern

Remote, AZ

August 2023 - December 2023

  • • Designed and implemented high-performance CUDA kernels for accelerating deep learning workloads, optimizing matrix operations, convolutions, and transformer models.
  • • Benchmarked and analyzed various AI accelerators, evaluating performance, power efficiency, and memory bandwidth for hardware-aware AI model optimizations.
  • • Collaborated with hardware engineers and AI researchers to co-design deep learning models optimized for edge computing and custom AI chips.

Automaton AI

Machine Learning Engineer

Pune, India

March 2021 - June 2022

  • • Led the deep learning backend development of an AI-powered auto-annotation tool used for object detection, segmentation, and classification, currently deployed in educational institutions and large enterprises.
  • • Integrated SSD, YOLOv3, YOLOv4, YOLOv5, RetinaNet, and Mask R-CNN into the tool’s API, streamlining pre-processing, training, evaluation, and deployment pipelines.
  • • Designed an optimized deployment pipeline, converting trained models to ONNX and further optimizing inference with TensorRT, significantly enhancing processing speed.