Close Search



Faculty of Engineering and Technology


About Me

I'm a Udacity Nanodegree graduate who loves developing and optimizing AI models in local environment/Edge devices without any need in sending data to cloud servers. My Latest project uses four different AI models to control mouse pointer movements with users “Eye Gaze”. I'm currently working on Software Optimization [SO] techniques like Intel's VTune Amplifier for finding hotspots in my application code.

Technical Skills

  • Proficient: C, Python, IoT, Edge computing, Machine Learning, Deep Learning
  • DL Frameworks: PyTorch, TensorFlow
  • Cloud Platforms: Intel DevCloud for Edge AI
  • Other Skills: OpenCV, Natural Language Processing, LaTeX, MySQL

HackerEarth’s Sigma-Thon 1.0 competition                                (July 2020 – HackerEarth)

To build a data-driven solution for retailers to innovate their retail channels with data models, recommendation engines, and much more. Grow your appetite for data-based solutions exponentially!

More info on the competition:

Have won Second Prize of 75$.

Project Title: “Smart video solutions for Fashion Store Optimization”

A Project which is a people counter system works on “Intel’s AI” which helps store managers to get insights on how well their store is performing.

More info on the project (GitHub Commits):


Intel Edge AI Foundation Course Scholarship                                 (March 2020 – Udacity)

Earned a place within the Top 850 out of 16450. Awarded the Intel Edge AI for IoT Developers

Nanodegree program from Udacity worth of Rs: 58,257/-


Machine Learning Challenge – STD drug Effective or not    (March 2020 – HackerEarth)

Achieved a Rank of 53 out of 3712, Created a model which has the ability to classify an STD drug is effective or not with 92.03% accuracy 



IEEE-CIS Fraud Detection | Kaggle                                                 (October 2019 – Kaggle)

Earned a place within the Top 79% in 6381. Fraud Detection model score: 88.43% 

Computer Pointer Controller:

  • Utilising Gaze Estimation model to control the mouse pointer of the computer.
  • The Gaze Estimation model requires three inputs: The head pose, the left eye image and the right eye image.
  • The total time taken by the models to perform inference is 8.1 seconds which has 0.49 FPS processing speed on Intel's i3 core processor.
  • This project has the ability to run multiple models in the same machine and coordinate the how of data between those models.


Smart Queuing Systems:

  • Deploying Person detection models in Edge devices by testing the hardware which best fits the client's requirements and their investments.
  • The models have been tested on the following hardware using Intel DevCloud:
    • CPU
    • VPU : Intel Neural Compute Stick 2 [NCS2]
    • IGPU
    • FPGA : Intel Arria 10 FPGA
  • Given different scenarios, the best hardware has been documented in proposed-template.pdf which meets the client's requirements and their investments to deploy AI models in Edge devices.