Rajesh Padhiary

Role :

• Group Lead for Advanced Engineering ,ADAS team .
• Responsibility for handling Innovation Tools, Software framework, Algorithm CPU/GPU optimization, Deep Learning AI algorithm & AI accelerator on chip for Next Generation sensor team
• Responsibility for leading rapid prototype-based project in Advanced Engineer team mainly for end-end prototype development , proof of concept , showcase next sensor performance ,early demos to customer for acquisition.
• Leading Advanced Topics mainly for new product features ,AI and Sensor algorithm to improve the performance of sensor
• Responsibility for front loading activity for Pre-development support mainly in framework development ,pre develop and test on sensor early samples also OEM research topic .
• Responsible for Software Architecture , Innovation Sprint , Project roadmap for Advanced Engineer project .
• Responsible for Project Management, Leadership, Hiring, Mentorship, road map, performance appraisal participation goal setting and feedback .
• Reviewed and revised software design to ensure technical compliance and proposal architecture improvements .
• Supervised developments teams ensuring 100% error-free code and design as well as on time delivery

Achievements :

• Create Center of Excellence in Advanced Engineer team for multiple field such Rapid Prototype Framework development, Visualization , Deep learning radar , algorithm optimization using CUDA & AI Perception on chip.
• Increase the team size from 1 member team to 40 member with multiple technical field .
• Proof of concept –”Semi Autonomous data labeling tool chain ” which reduces manual labeling by a factor of 80%. Due to this CBS(Continental Business solution ) series project started also it is awarded one of the best publication in Artificial intelligence robotics conference.
• Deep Learning Radar Road Edge detection which able to achieve maximum performance beyond 200 m(Highway) and up to 90 m (in curve Scenario) .
• Rapid prototype and quick demonstration of functionality of next generation sensor which help multiple acquisition ,pre-development and business decision .
• Leading and completion of Self driving pre –development project(L2/L2+ system) with tier-1 OEM due to this base project has been started .
• Leading algorithm optimization activity using AI accelerator or compression or CUDA optimization which help to demonstrate the perception algorithm /sensor performance in vehicle .

Major Project :

Semi Auto data Labeling Tool:

Proof of concept –”Semi Autonomous data labeling tool chain ” which reduces manual labeling by a factor of 80%. Due to this CBS(Continental Business solution ) series project started also it is awarded one of the best publication in Artificial intelligence robotics conference.

Transfer Learning :

• Automated driving make high demands to environment perception with multiple sensors .Multiple Sensor Fusion is needed for highly reliable detection performance
• Transfer Learning tools used to transfer label data from sensor to another .
• Semi-Auto Labeling Radar Clusters using Image Segmentation ›Radar cluster classification and segmentation using Deep Learning

Road Edge using Deep Learning :

• Road boundary detection to max range (above 200m ) is a key requirement for EuNCAP
• Low Level Fusion by means of Deep Learning Radar.
• Approach : Radar detection segmentation , Clothoid based road edge , Key point based ,Fusion between camera and radar detection for accurate ego lane road edge

TEST & VALIDATION TOOLS FOR CEM (COMPREHENSIVE ENVIRONMENT MODEL @ ADAS)

• Design & development of tool chain addressing complete workflow from data recording to KPI generation for testing Comprehensive environment which help to start series project
• Tool chain support both real and synthetic data , provides feature to tag scenario such as city ,highway ,tunnel etc ,Test data management and finally automatic KPI generation for quantitative testing

Next Generation Sensor Visualization :

• Visualization tools enable the developer to visualize, debug and analyze sensor data, sensor performance, algorithm output and algorithm behavior.
• Pluggable Architecture , Multi Platform Support ,Real/Synthetic data support ›Visualization Feature Support : Sensor data (Radar , Camera, lidar ) , Video wall , 4D power graph ,Traffic Participant , Freespace road boundary ,3D animation model –classification ,road boundary , localization & mapping
• Quick demonstration of functionality of next generation sensor which help multiple acquisition ,pre-development and business decision

Pre –Development OEM Project :

• Design & Development of end-end software development based on enhanced Communication Abstraction Layer (eCAL) architecture as middleware which provides scalable and high performance interprocess communication .
• Rapid prototype of next generation sensor such as radar (short & long range) ,camera (mono ,surround view ) ,perception Algorithm (classical & AI based) to demonstrate 360 degrees Comprehensive environment model for ADAS functionality such as ACC & Motion control, Highway Assist/Hands-off Highway etc
• Software needs to support offline/online and vehicle demonstration for NCAP and Autobahn scenario

ADAS ALGO camera and Radar OPTIMIZATION:

• ADAS Perception algorithm CPU & GPU optimization to demonstrate the functionality on vehicle ›Generic Soft Image Signal processing (ISP) Algorithm and optimization using CUDA to demonstrate real time on vehicle for Next Generation Camera .It helps to test the perception algorithm on pc before porting to the hardware .
• Algorithm optimize from 500ms to 3ms on NVidia graphic card
• 360 Pedestrian Detection for surround view camera algorithm optimize using CUDA from 33 ms to 15 ms on Nvidia Drive PX2
• Radar based Grid Processing freespace detection algorithm optimized using CUDA ,c++ and OpenGL

Precise Localization & Mapping :

• Precise localization is a key prerequisite for assisted and automated driving for ACC ,Enhanced EBA , freespace detection etc..
• Radar Map generation and localization using SLAM
• Precise mapping of Sensor detection over map data

Perception Algo on Chip

• Evaluation of multiple hardware chip on how to run AI based Perception Algo

• Led team of engineers in providing design and engineering solutions to various clients and market sectors.
• Leading AGCO Combine Harvester project for machine control for all harvesting functions
• Design of CEA Gen Tool which is used for auto generation of source code

Role & Achievements:

• Design and Development of ACHD 2.0 (Advanced Video Coding High Definition) plugin in P2Viewer Plus used to play back, copy clips, display, upload and editing metadata, easy editing and perform other operations supporting AVCHD 2.0 format
• Design and development of AVCHD parser and AVCHD writer.

• Design & Development of EBA which is used for digital signage application along with benchmarking results on selected Intel hardware running different plugin such as multiple content as video, Audio, Images, RSS, Text Tickers, Adobe Flash, Security, Medical and widget (clock, face detection etc.).

• Design & development of HMMP encoder tool for 3D/2D video files. This tool uses H.264/AVC (Advanced Video Codec) and AAC (Advanced Audio Codec) to encode source video and audio respectively and thereby provides the high-quality visual image
• Design & development of Distributive encoding for 3D/2D feature
• Design & development of Subtitle graphics stream feature

• Design & development of static analysis, memory leak & generic optimized building tools for development & testing of mobile software based on Symbian platform

• Design of Development of speech recognition software for voice navigation using Nuance VoCon ASR (Automatic speech Recognition) SDK

• Design and Development of CENTRAL SUBSCRIBER DATA ENTITY (CSDE) in CRBT.

• Design & development of e-mail client application using SMTP (Simple Mail Transfer Protocol) based on S60 Symbian platform

Module: Central Transceiver Unit (CTRU) i.e. BASESTATION in WLL.

• Bug Fixing & Development of Central Transceiver Unit in Air loop system