Vipul Parashar

CEO

Pune, Maharashtra, India17 yrs 11 mos experience
Most Likely To SwitchAI Enabled

Key Highlights

  • 16 years of experience in software engineering.
  • Expertise in AI and Deep Learning technologies.
  • Proven track record of innovation in video compression.
Stackforce AI infers this person is a software engineer specializing in AI-driven video processing solutions.

Contact

Skills

Core Skills

AiDeep LearningEmbedded SystemsMemory ManagementMiddleware DevelopmentAndroidNetwork ProgrammingPacket ManipulationDlnaMultimedia ApplicationsMessaging Frameworks

Other Skills

AlgorithmsCC++CNNCUDAComputer VisionDebuggingDesign PatternsDevice DriversGraphics ProgrammingJavaJavaScriptLinuxMobile DevicesMultithreading

About

Highly accomplished and results-driven software engineering professional with 16 years of experience in designing and developing cutting-edge solutions across diverse domains, including video codecs, optical flow, and deep learning-based algorithms for video compression. Proficient in AI and Deep Learning technologies, with a proven track record of driving innovation and delivering impactful results.

Experience

Nvidia

Senior System Engineer

Jul 2015Present · 10 yrs 8 mos · Pune Area, India

  • I specialize in advancing the frontier of video restoration, super-resolution, and compression by applying cutting-edge AI and deep learning techniques.
  • My work spans:
  • Developing and training autoencoder and diffusion models for video restoration and resolution enhancement.
  • Leading R&D on CNN and Transformer-powered video saliency prediction, resulting in two filed patents.
  • Building CNN-based prefiltering models to boost video compression quality.
  • Creating an end-to-end AI surrogate model for next-generation video compression pipelines.
  • Pioneering CUDA-accelerated solutions such as temporal filtering and MBTree, achieving 15% greater HEVC coding efficiency (1 patent granted, research presented at Video Mile High Conference).
  • Co-inventing a patented frame rate up-conversion (FRUC) system using optical flow (1 patent granted, 2 filed).
  • Enhancing raw optical flow quality with CV-CUDA based pre/post-processing techniques.
  • Engineering public APIs, device drivers, and core components for Nvidia optical flow hardware and NVENC driver stacks.
AIDeep LearningVideo CompressionCUDAOptical FlowDevice Drivers

Samsung india software center noida

4 roles

Senior

Sep 2013Jun 2015 · 1 yr 9 mos

  • Have worked on memory manager (Garbage collector), task Manager and graphics foundation layer of compact virtual machine (CVM) for embedded devices (BD-Player DTV).
  • Design and develop porting layer to port embedded virtual machine on X11 windowing system and Cairo Graphics library.
  • Design and develop Heap Memory manager (Garbage collector) for VM.
  • Design and Develop Task Manager for VM.
  • Design and Develop Porting layer of VM graphics profile for embedded devices.
  • VM commercialization issues debugging for embedded devices.
  • Design and Develop JVM native interface porting layer that supports X11 window system and Cairo 2D Vector graphics.
Memory ManagementTask ManagementGraphics ProgrammingEmbedded Systems

Senior Software Engineer

Sep 2011Aug 2013 · 1 yr 11 mos

  • Middleware designing and development to port Android on Linux based O.S. tizen (Our team developed a middle ware between Android Application and a Linux based Operating system Tizen, to facilitate Android Application on Tizen. We ported Android FW (framework) and Android HAL (Hardware abstraction layer) over Tizen.)
  • Designed and developed Android boot process on Tizen(Linux based) operating system.
  • Have good understanding of Binder Inter process communication.
  • Tizen O.S supports eglibc,While Android support bionic libc. Our team designed and developed customaries DVM (Dalvik virtual machine), Linker and loader for Android native Application (NDK) support on Tizen O.S.
  • Designed and developed Android framework and Android hardware abstraction layer to port Android Wi-Fi, GPS, and on tizen operating system.
  • Development Tools : C , C++, Linux
Middleware DevelopmentAndroidLinux

Senior Software Engineer

Mar 2011Sep 2011 · 6 mos

  • Development of network parsing tool for TCP, UDP, IP protocol transform/Drop the packet data.
  • 1. To capture packets (TCP, UDP, IP) in the network and manipulate packet as per user requirement and forward it to required destination.
  • Development Tools : C , VC++, Windows
Network ProgrammingTCP/IPPacket Manipulation

Senior Software Engineer

Sep 2010Feb 2011 · 5 mos

  • Development of DLNA application Like DMC, DMS, DMP and Maintaining of DLNA stack (DLNA aims to promote wired and wireless interoperable networking of PCs, CE (consumer electronics) and mobile devices)
  • 1. Participated in the development of DLNA stack, my core responsibility was debugging and the issues in the stack. 
  • 2. Developed search capabilities and SSDP update (updating on DMC (controller) side when any update on the DMS (server) side), in the DLNA stack.
  • 3. Developed DMP (Digital Media Player), which is similar to DMC but it has its own player for playing of multimedia data.
  • Development Tools : C , C++, Windows
DLNADebuggingMultimedia Applications

Aricent technology limited

Software Engineer

Jul 2008Aug 2010 · 2 yrs 1 mo

  • Long Term Evolution: Messaging Frameworks, Memory Manager Handling ,UE
  • I worked for memory manager, Messaging framework, DCT(data collection tool of UEs),
  • 1. Memory Manager static library: In a Real Time System to avoid the run time memory allocation/de-allocation, private buffers are created for processes/threads. This library provides the interface to achieve that.
  • 2. Messaging Framework library: It was POSIX message queue based IPC library toprovide interface for processes/threads to exchange messages.
  • 3. Enhanced the “timer library:”: which provides the interface to the processes/threads to start guard timers.
  • 4. DCT (Data Collection Tool) SW: This SW refers to a tool for directly collecting, displaying and storing information, such as measurement items and the internal state of the equipment. This is enabled by directly connecting the client terminal (DCT-PC) to the eNodeB base station equipment.
  • 5. Implemented the Serializer/De-serializer for certain SOAP-XML messages. The Serializer function converts the C message to XML message and De-serializer does the reverse conversion.
  • 6. Implemented some UE control APIs, providing the function to fetch the “number of active users (UE) per cell” and for the “number of users undergoing handover procedures and related source/target cell information”.
  • 7. Worked on the X2AP Protocol that is Protocol to create interfaces between Two eNodeBs.
  • Duration: Sep 6,2010 - Dec 2010
  • Multi Party Conference Chat Server
  • 1. The project was aimed to chat among multiple clients through single server in multiple sessions. Socket programming used in this project.
Memory ManagementMessaging FrameworksReal-Time Systems

Aricent

Trainee

Jan 2008Jul 2008 · 6 mos

  • My job responsibility was to test the 3G call in femtocell.

Education

National Institute of Technology Karnataka

MCA — computers

Jan 2005Jan 2008

Stackforce found 100+ more professionals with Ai & Deep Learning

Explore similar profiles based on matching skills and experience