Levin Dabhi

Levin Dabhi

Machine Learning Engineer

Ceremorphic Inc, Hyderabad

Hello, I’m a Machine Learning Engineer working at Ceremorphic Inc in the compiler development team. Our team works on the R&D of Machine Learning compilers for low power AI accelerators. Prior to this, during my academics, I worked on research problems in the area of image and video generation. Apart from my day job as a Machine Learning Engineer, I like to play with generative models.

I am most skilled at Python, C++ and deep learning framework PyTorch. I have experience with technology like C, Java, DL framework Jax, TensorFlow and Keras, HTML, CSS, Web framework Flask, SQL, Google Cloud Platform, AWS and Azure.

I like solving a mathematics puzzle, playing-watching video games, and playing-watching Garba apart from above things. If you are interested to know more about me, want to discuss any of my work, want to discuss any new idea, or want to discuss any ongoing esports tournament say hello on Twitter.

Posts

Generating Amongus Characters

Results of training stylegan2-ADA with few images
Feb 7, 2021 2 min read
Generating Amongus Characters

Generating different styles from stylegan

Results of blending various models in stylegan/stylegan2
Oct 12, 2020 4 min read
Generating different styles from stylegan

Experience

 
 
 
 
 

Machine Learning Engineer

Ceremorphic Inc

May 2021 – Present Hyderabad, India
 
 
 
 
 

Computer Vision Intern

Spyne

Jan 2021 – May 2021 Gurgaon, India
  • Developed an algorithm to perform virtual try-on for different garments with high-resolution output images by employing SoTA deep learning models.
  • Worked on background removal and replacement from images of cars, models, and shoes. Developed a highly scalable and efficient training pipeline of background removal for other categories.
 
 
 
 
 

Data Science Intern

Indian Naval Ship (INS) Valsura

Jun 2020 – Jul 2020 Jamnagar, Gujarat
  • Worked on Exploratory data analysis of Ship engine data and developing a predictive maintenance system for fault classification in that engine.
  • Worked on various imputers, and dimensionality reduction techniques like finding different types of Correlations, and Principal Component Analysis. Used various Kernel Density plots for fault classification.
 
 
 
 
 

Machine Learning Intern

Infivolve Inc.

Sep 2019 – Jun 2020 Ahmedabad, India
  • Worked on developing a virtual coach for Basketball and Lawn Tennis sports.
  • Developed algorithm that tracks player and ball from the video in real-time using deep learning techniques. Reduced the inference time of the tracking algorithm by a factor of 2.4
  • For Basketball sport, Worked on the analysis of various parameters like speed and angle of a ball from videos. For Lawn Tennis, Worked on analysis and plotting of various parameters like bounce point, hit point, and speed of the player. Developed court mapping technique that maps players and ball in 2D for both sports.
 
 
 
 
 

Research Assistance

Nirma University

Apr 2019 – Jun 2019 Ahmedabad, India
  • Research project was funded by the Board of Research and Nuclear Sciences (BRNS)-Department of Atomic Energy.
  • Worked on optimal trajectory finding of robotic hand by solving inverse kinematics to reach a particular point which is damaged in Tokamak using Deep Reinforcement Learning. Developed a model that predicts the trajectory of the robotic hand in simulation. The simulator was developed in the blender tool as per requirement.
  • Implemented policy gradient algorithms like Proximal Policy Optimization(PPO) and Deep Deterministic Policy Gradient(DDPG) with a deep neural network as a function approximator.

Contact