Elastic-NeRF
nerf
elastic
wip
Making NeRFs even more compact and efficient using an elastic architecture (and no additional training cost)!
Jan 9, 2024
Bio
I’m a first year PhD student in Systems Design Engineering at UWaterloo’s Vision and Image Processing Lab. I’m advised by Prof. Alexander Wong and Prof. Yuhao Chen and supported by an NSERC CGS-D award. My research has focused on projects at the intersection of efficient ML (specifically applying neural architecture search to NeRFs and vision transformers), embedded systems (latency prediction for on-device inference), and robotics (building simulation pipelines for synthetic data generation, training embodied morphologies, and dexterous manipulation). I’m interested in building embodied systems that can efficiently learn to interact with the world in an open-ended manner.
I previously graduated from the University of Waterloo with a MASc in Systems Design Engineering and a BASc in Mechatronics Engineering with an Option in Artificial Intelligence. I have industry experience as a software engineer and AI researcher working across the full systems stack including sensor processing and communications infrastructure on self-driving cars, teleoperation & imitation learning for humanoid robots, simulators for robot grasping, real-time video analytics applications for embedded microcontrollers, safety-critical bootloader firmware for aerospace products, silicon validation for AI accelerators, and web development for e-commerce.
Recent Projects
Check out my projects page for more!
No matching items
Industry Experience
Nuro.ai | Software Engineering Intern
Brought up hardware accelerator, implemented efficient digital signal processing algorithms (in C) for radars on self-driving delivery robot, and developed testing architecture for radar HIL simulation chambers.
BRAIN Lab, University of Waterloo | Undergraduate Researcher
Developed a rigid-body grasping simulator in PyBullet and used synthetic demonstration data to train a planner for robotic assembly tasks.
Nuro.ai | Software Engineering Intern
Developed onboard and embedded communications infrastructure for radar driver module on self-driving delivery robot. Improved sensor data throughput and firmware upgrade speeds.
Arcturus Networks | Embedded Machine Vision Developer Intern
Designed real-time, multithreaded video analysis application in C++ for low power embedded platform. Improved multi-object tracking algorithms and developed a library to accelerate CPU-based inference using ARM NEON intrinsics.
Curtiss-Wright Defense Solutions | Embedded Software Engineering Intern
Implemented, validated, and released features for U-Boot bootloaders. Drove debug and validation efforts across VxWorks & Yocto Linux BSPs, BIOS, and bootloader products to meet release schedule.
Centre for Operational Research and Analysis, DRDC | Research Engineer
Prototyped tool for CANSOFCOM to optimize troop schedules using genetic algorithms.
NXP Semiconductors | AI & Vision Processor System Level Design Intern
Developed an internal tool to simulate DMA performance of NXP’s accelerated vision processor for ADAS applications. Analyzed results and waveforms to identify bugs in SystemVerilog RTL resulting in 2.73x increase in DMA bandwidth.
Home Depot Canada | Junior Front End Web Developer
Developed features for an internal front-end component library to assist migrating the Home Depot Canada website to React.
Selected Publications
Saeejith Nair, Yuhao Chen, Mohammad Javad Shafiee, Alexander Wong. NAS-NeRF: Generative Neural Architecture Search for Neural Radiance Fields. arXiv:2309.14293, 2023. NewInML workshop @ NeurIPS, December 2023.
Saeejith Nair, Javad Shafiee, Alexander Wong. DARLEI: Deep Accelerated Reinforcement Learning with Evolutionary Intelligence. arXiv:2312.05171, 2023. Conference on Vision and Intelligent Systems (CVIS), December 2023 (Oral).
Saeejith Nair, Chi-en Amy Tai, Yuhao Chen, Alexander Wong. NutritionVerse-Synth: An Open Access Synthetically Generated 2D Food Scene Dataset for Dietary Intake Estimation. arXiv:2312.06192, 2023. NewInML workshop @ NeurIPS, December 2023.
Alexander Wong, Saad Abbasi, Saeejith Nair. TurboViT: Generating Fast Vision Transformers via Generative Architecture Search. arXiv:2308.11421, August 2023. NewInML workshop @ NeurIPS, December 2023.
Chi-en Amy Tai, Matthew Keller, Saeejith Nair, Yuhao Chen, Yifan Wu, Olivia Markham, Krish Parmar, Pengcheng Xi, Heather Keller, Sharon Kirkpatrick, Alexander Wong. NutritionVerse: Empirical Study of Various Dietary Intake Estimation Approaches. Multimedia Assisted Dietary Management workshop @ ACM Multimedia Conference, September 2023.
Alexander Wong, Yifan Wu, Saad Abbasi, Saeejith Nair, Javad Shafiee. Fast GraspNeXt: A Fast Self-Attention Neural Network Architecture for Multi-task Learning in Computer Vision Tasks for Robotic Grasping on the Edge. Neural Architecture Search workshop @ CVPR, June 2023, pp. 2293-2297.
Chi-en Amy Tai, Matthew Keller, Mattie Kerrigan, Yuhao Chen, Saeejith Nair, Pengcheng Xi, Alexander Wong. NutritionVerse-3D: A 3D Food Model Dataset for Nutritional Intake Estimation. arXiv:2304.05619, Apr 2023. WiCV workshop @ CVPR, June 2023.
Alexander Wong, Javad Shafiee, Saad Abbasi, Saeejith Nair, Mahmoud Famouri. Faster Attention Is What You Need: A Fast Self-Attention Neural Network Backbone Architecture for the Edge via Double-Condensing Attention Condensers. All Things Attention workshop @ NeurIPS, December 2022.
Saeejith Nair, Saad Abbasi, Javad Shafiee, Alexander Wong. Maple-Edge: A Runtime Latency Predictor for Edge Devices. Embedded Vision workshop @ CVPR, June 2022 (Oral), pp. 3660-3668.
Xueyang Yao, Saeejith Nair, Peter Blouw, Bryan Tripp. Inferring symbols from demonstrations to support vector-symbolic planning in a robotic assembly task. hal-03041290, November 2020. SMILES (Sensorimotor Interaction, Language and Embodiment of Symbols) workshop @ ICDL, November 2020.
Pronunciation guide
My first name is pronounced like sigh-jith (IPA: /ˈsaɪ-dʒɪθ/).