Nitthilan Kannappan Jayakodi

Nitthilan Kannappan Jayakodi

PhD in Artificial Intelligence / Machine Learning

Washington State University

Biography

I am Nitthilan, pursuing PhD in AI/ML for Video/Image/3D Graphics. Interested in image and video based motion capture using pose based estimation (SMPLX, FrankMocap, SMPLify-X), clothing and hair 3D model representation using Implicit Neural Representation (NeRF, NSVF, Neural Body) and Deep generative models (GAN, VQVAE) .

  • 12 years of software product development experience with wide exposure in building end to end systems from embedded to web technologies.
  • Passionate about developing optimized algorithms and exploring different domains. Developed video applications for conferencing, broadcast, and storage. Worked on ME/RC for H264/MPEG and developed plugin/webRTC/Html5 based video endpoints

Current Focus Area:

  • Extracting facial pose and full body motion capture (dance steps, sport stances, exercise and fitness poses) data from monocular video (YouTube, TikTok) using SMPLX prior (FrankMocap, SMPLify-X, VIBE)
  • Controllable cloth, hair attribute and facial mocap of dynamic humans using SMPLX anchored latent codes (NeuralBody, FrankMocap, HumanParser)
  • Accelerating learning of the 3D Neural Scene representation using smaller neural nets and latent representation at a voxel level [NeRF, NSVF, KiloNeRF].

PhD Research: My research is at the intersection of Machine Learning (ML) and Computing Systems (Sys). The overarching theme of my research is to bridge these two areas. Specifically, I’m working towards the vision of Edge AI to efficiently deploy AI solutions for emerging applications (e.g., robotics, selfdriving cars, augmented/virtual reality, and smart health) on edge platforms that are constrained by resources (power, compute, and memory).

(a) Developed a novel hardware-aware design and optimization framework to trade-off power, performance, and accuracy for performing inference using deep neural networks.

(b) Studied effective instantiations of this framework for different applications - Convolutional neural networks for image classification, Graph convolutional networks for 3D computer vision, and Generative adversarial networks for image manipulation tasks.

(c) Published papers at top-tier venues including three journal papers (ACM/IEEE Transactions on CAD and Embedded Computing Systems) and four conference papers (AAAI, DAC, DATE, ICCAD). One journal paper is under review at Journal of Artificial Intelligence Research.

Interests
  • Extracting facial and full body pose from monocular images and videos
  • Neural scene representation for fast and high-quality free-viewpoint
  • Generative Models - GAN, VAE - Video/Image/3D Graphics
  • Bayesian Optimization Approaches
  • Artificial General Intelligence
Education
  • PhD in Artificial Intelligence, 2022 (Pursuing)

    Washington State University, 3.95/4 CGPA

  • BE Electronics and Communication, 2004

    College of Engineering, Anna university, , 8.91/10 CGPA

Skills

PyTorch, Keras (DL Framework)

90%

C, C++, Embedded Systems

90%

Full Stack (MEAN stack)

90%

Video Codec (SVC, H265, H264, MPEG2/4)

90%

Experience

 
 
 
 
 
Senior Staff Developer (Technical Architect)
May 2012 – Aug 2017 Bangalore, India

Responsibilities include product design, implementation, testing, and delivery:

  • Successful deployments of Polycom Telepresence solutions @ major institutions and companies
  • Video Conferencing plugin for chrome 64 bit using PPAPI
  • RTP, RDP and HTML5 based Content Collaboration platform between MCU, Lync and Browser endpoints
  • Automation Platform for bringing up Infrastructure for Video as a service on VMWare
  • Platform Director: Life cycle management of virtual instances on VMWare using viJava
 
 
 
 
 
Lead Engineer / Senior Engineer
Aug 2007 – May 2012 Bangalore, India

Responsibilities include product design, implementation, testing, and delivery:

  • Managed small teams of size 4-5 people
  • Successfully delivered to customers like Sony, Harman, Pioneer, Amagi, Samsung, Huwawei, Mango Systems, Parrot Systems,
  • Mpeg2 (MP@HL) HD(1080i@30fps) Encoder/Transcoder for Broadcast on Netra (IVAHD Accelerated)
  • MPEG4 Simple Profile (SP) HD(720p@30fps) Encoder for Smart Phones on OMAP34xx (Arm+DSP+IVA)
  • Mpeg4 SP D1 (480p@30fps) Encoder for Portable Media Player and Recorder on C64x+ DSP (OMAP3430)
 
 
 
 
 
Senior Engineer R&D Video Codec
Aug 2004 – Aug 2007 Bangalore, India
  • Managed remote teams for Stretch Inc in H264, MPEG4 codec development
  • Managed design development and delivery for codecs for Canon Camera Systems
  • Developed codec on Intel SSE tecnology for MPEG2, H264 codecs. Video Encoders.

Recent Publications

Quickly discover relevant content by filtering publications.