Nathiyaa Sengodan
Senior Engineering Manager - Platforms and Compilers at MulticoreWare Inc- Claim this Profile
Click to upgrade to our gold package
for the full feature experience.
-
English -
-
Tamil -
Topline Score
Bio
Experience
-
MulticoreWare Inc
-
United States
-
Software Development
-
300 - 400 Employee
-
Senior Engineering Manager - Platforms and Compilers
-
May 2023 - Present
-
-
Engineering Manager
-
Mar 2022 - Present
- Manage projects related to NN kernel / CV algorithm optimization for target DSP & AI/DL accelerators.- Manage projects related to compilers (LLVM, MLIR, etc)- Focused on expanding the team with right skillset while maintaining the productivity and quality of the deliverable
-
-
Senior Lead
-
Jan 2021 - Mar 2022
- Product Development : HuBe.ai (First of its kind Human Behavior Intelligence (HBI) Platform - AI/API/SaaS)- Human pose estimation and it's use cases- Human activity recognition- Body language identification- Kernel development for NN architectures for AI accelerator- Algorithm optimization for better throughput performance
-
-
Technical Lead
-
Jul 2020 - Jan 2021
- Drowsiness and distraction detection pipeline development for focus monitoring- Drowsiness detection | Eye state monitoring, Yawn detection, Microsleep detection- Distraction detection | Head pose estimation & Eye gaze estimation- Porting focus engine modules (Distraction & Drowsiness) for target hardware to achieve real-time performance- End-to-end ML/CV based Driver Monitoring System- Senior Assist application | Porting the application onto the target DSP- HuBe.ai Product development | Focus Engine, HAAPI & BOLA Show less
-
-
Senior Software Engineer
-
Jul 2016 - Jun 2020
- Vehicle detection using SSD and vehicle tracking algorithm implementation to count number of vehicles on the road- Real time Vehicle detection based on YOLO tiny architecture. Implemented and optimized YOLO tiny inference kernels for the target DSP using different DSP specific intrinsic, Data tiling and DMA - Quantization of Inception-V3 architecture to INT8. Porting of all the layers in the architecture to the target DSP using DSP specific intrinsic, Data tiling and DMA. Optimization of the layerwise kernels to achieve actual CPP that's close to theoretical CPP- Design & implementation of a NN model that is capable of localizing and recognizing various traffic signs on the road- Multi-Object detection using CNNs to localize and recognize Vehicle, Pedestrian and Traffic Signs- Design and development of multi-object tracking algorithm using CNN and OpenCV techniques- Porting Image captioning pipeline for the target DSP- LipSync pipeline porting onto a Intel platform using OpenVINO framework- Object detection pipeline porting onto Movidius Stick | Throughput performance 5 FPS- Optimized numerous NN kernels for a popular AI accelerator hardware including Filter 3D, AvgPool 3D, Uniform distribution, Asin & Acos, GRU Stage 1 & 2, etc., Show less
-
-
Software Engineer
-
Jul 2014 - Jun 2016
* CUDA Parallel programming* CNN classifier application for Face and vehicle classes* CNN Detection application for Face and Vehicle classes* Human Action Recognition pipeline* Familiar with Caffe & Torch frameworks
-
-
Education
-
Coimbatore Institute of Technology
Bachelor of Engineering (B.E.), Computer Science