I'm Arjun A B!
AI enthusiast with hands-on experience in various AI models. I work on the development of
industry-leading high-performance and power-efficient GPUs. I am committed to broadening
knowledge and leveraging expertise to contribute to game-changing advances in the field. My
collaborative nature, adaptability, and meticulous attention to detail set me apart. I am looking for
impactful projects where I can contribute expertise and drive innovation.
Email |
LinkedIn |
Github
|
|
Experience
SandLogic Technologies Pvt. Ltd.
Edge AI Developer (Sept 2022 – Present)
- Automatic toll collection using Deep learning with Data collection of 4 different classes of vehicles containing more than 3500 images for Indian vehicles. Data analysis and data cleaning for the Implementation of toll. Model accuracy of 97.2% was achieved using YOLO.
- Optical character recognition (OCR) with image quality assessment. A brief study of edge detection using different
filtering techniques. Dot-printed character recognition for multiple day-to-day used products. Filtered and extracted
text yielded 92% of accuracy.
- Identification of abnormal ECG signals using a low-power microcontroller [Tiny ML].
- Detection of various arrhythmias with the lowest latency of 36.424 ms and a model accuracy of
98%.
- The dataset included 1200 datapoints from 20,000 patient records.
- Implemented on a microcontroller STM32 Cortex-M4 with 256 KB of flash memory.
- Deep learning accelerator was implemented on the Chromite-H core system-on-a-chip (SoC).
- Incorporated Deep Learning Accelerators into the RISC-V SoC Incore Chromite-H core, utilizing
Zynq Ultrascale+ MPSoC ZCU104 FPGA.
- To enable the SoC and Accelerator to share the DDR Memory Controller, an integrated Multi-
Port Memory Controller (MPMC) design was developed.
- Accumulated practical understanding of the various AXI Infrastructure Architectures and the AXI
Protocol.
- Timing analysis, placement & routing, and a variety of built-in Vivado methods were used to
optimize the FPGA design.
- Ran a range of DL models and adapted Accelerator's runtime and kernel drivers to the RISC-V
Linux kernel.
- Implementation of Deep Learning Accelerator on PolarFire SoC FPGA Icicle Kit. Configuration of the RISC-V
processor using MSS configurator tool. Conducted a complete study and understanding of the Libero SoC tool and
Polarfire SoC.
- Design and implementation of high-performance CSI2 MIPI camera system on ZCU104
- Designed a dedicated MIPI Circuit with specialized filters to ensure flawless image data flow
- Integrating VCU (Video Codec Unit) to Processes frames and transmit them to gstreamer.
- Ensured seamless frame handling, guiding frame acquisition, configuration, and integration with
gstreamer for video streaming and processing.
- Created unique IP blocks to enhance versatility and performance, addressing specific needs.
- Developed a tailored operating system to provide a robust foundation for camera interaction and
peripheral support.
- Utilized Xilinx Vitis to optimize application deployment within the FPGA by bridging hardware
and software.
SandLogic Technologies Pvt. Ltd.
Edge AI Intern (Apr 2022 – Aug 2022)
-
Understanding the basics of AI and Designing deep learning neural nets
- Data collection and collation for training the model for vehicle detection.
- Data annotation and labelling using make sense.ai
- Training, Validation, and Testing for NMIST.
- Inference check.
- Autopilot system for cars with Advanced Driver Assistance System.
Personal Projects
Smart switches using Node MCU (08/2021)
- Best workshop project award.
- Controlling the switch through a mobile app protected by a Google firewall.
- It could be used in a variety of crucial situations where human assistance would be difficult.
Diabetes prediction using PIMA dataset (02/2022)
- Diabetes prediction based on the PIMA Indian dataset.
Implementation of E-Toll using Deep Learning (12/2021 - 06/2022)
-
Best presenter award at ICAMET Conference.
- CNN-based automatic vehicle detection.
- Vehicle classification was archived swiftly, which reduced congestion at the toll booth.
- A web application-based graphical user interface (GUI) was created to monitor the progress of
vehicle classification and tax implementation.
Education
Vidyavardhaka College of Engineering – Mysore, India
Bachelor of Engineering, Electronics and Communication; Cumulative GPA: 8.5/10 (May 2022)
Technical Skills
- Technical: Field-Programmable Gate Arrays (FPGA), Digital logic design, Systems Design, High Performance Computing, Computer architecture, Hardware-software co-design.
- AI concepts: Artificial Intelligence, Machine Learning, Neural Networks, Generative AI, Large Language Models
- Languages: Python, C/C++, Bash
- Hardware Languages: Verilog, System Verilog
- Design Tools: Xilinx Vivado, Vitis HLS, Microchip Libero, Matlab & Simulink
- Protocols: UART, I2C, SPI
Achievements
- Best Paper Presenter Award at Second International Conference on Advances in Management, Engineering & Technology (ICAMET-2022) (June 2022)
Certificates
- Programming for Everybody (March 2021 - April 2021)
- Intel Distribution of OpenVINO Toolkit (April 2021)
- Second International Conference on Advances in Management, Engineering & Technology (ICAMET-2022) (June 2022)
- 2022 ACM/IEEE TinyML Design Contest at ICCAD
|
|