• AD
  • Akhil Devarashetti

    Deep Learning Engineer

    akhilez.ai@gmail.com

Bio

I am a Deep Learning Engineer with background in Software, Master's in Artificial Intelligence.

I am obsessed with Neural Networks šŸ§ , Evolution šŸ§¬ and Astrophysics šŸŖ

Timeline

2021 Nov

Machine Learning Engineer ||| - R&D

ACV Auctions

R&D of various Deep Learning based solutions for car inspections.

2021 May

Computer Vision Engineer

BioTrillion

R&D of pupillometry using Deep Learning techniques like Semantic Segmentation, Detection, etc.

2020 May

Developer

VNDLY

NLP Tasks - Semantic Text Similarity based on BERT, segments highlighter
Technologies: PyTorch, TensorFlow, Django, React

2019 Aug

Masterā€™s in Artificial Intelligence

University of Cincinnati

Specializations: Computer Vision, NLP, Reinforcement Learning and Complex Intelligent Systems

2018

Full-Stack Engineer

Aviso AI

Reduced the ML cloud cost by 60%
Technologies used: Django, Kubernetes, AWS, Linux, Puppet, Vue.js

2015

Part-Time Developer

Keshav Memorial Institute of Technology

Developed apps for the collegeā€™s operations like Student Feedback and Gate-Pass System

2014

Bachelor's in Computer Science and Engineering

Keshav Memorial Institute of Technology

My first adventures with programming

Coding Activity

I absolutely ā¤ coding! Each box below represents the amount of coding on that day of the year. Darker box = lot of code

From GitHub @Akhilez

Loading ā€¦

Public Talks

Overview on Reinforcement Learning

July 15, 2021

I spoke at this July's CincyML meetup where I introduced Reinforcement Learning in a very approachable way. I used PyTorch and other libraries to train my machine to play games like GridWorld and even Super Mario.

reinforcement learningpytorchgym

Research Works

  • DisCease: Evolving Social Distancing And Herd Immunity

    Abstract: I tried to create a model that simulates the spread of a disease that does not have a medicine. I observed the effects of varying parameters, then introduced a concept of deflections which mimic social distancing and social gatherings. I finally made an attempt to evolve these deflections based on a performance metric.

    ReadRun
    DisCease: Evolving Social Distancing And Herd ImmunityImage
  • A Brief Survey of Model-Free Deep Reinforcement Learning

    Abstract: Deep Reinforcement Learning is a branch of machine learning techniques that is used to find out the best possible path given a situation. It is an interesting domain of algorithms ranging from basic multi-arm bandit problems to playing complex games like Dota 2. This paper surveys the research work on model-free approaches to deep reinforcement learning like Deep Q Learning, Policy Gradients, Actor-Critic methods and other recent advancements.

    A Brief Survey of Model-Free Deep Reinforcement LearningImage

Hackathons

  • RevolutionUC

    22nd February, 2020 | Top 5 for Alster AR

    RevolutionUC is a hackathon conducted by ACM at University of Cincinnati with roughly 400 participants.

  • NeuralHack

    16th November, 2017 | Won 3rd

    NeuralHack is an India-wide hackathon with approximately 13,000 participants conducted by Virtusa

Projects

  • Deep Learning

  • Mobile

    • Syllabus

      Syllabus

      An android app that provides Engineering syllabus. Aimed for students to manage and keep track of their studies. This became quite popular at my college.

      androiddjangomysql

      2016

    • MedicWhizz

      MedicWhizz

      An educational service for medical students (Plab). Provides practice quizzes for medical students in the form of a beautiful mobile app. Students can visualize their learning progress.

      flutterfirestore

      2019

  • Web

    • Student Feedback

      Student Feedback

      A comprehensive web application for universities to conduct faculty (and other) rating feedback sessions based on various teaching aspects. It collects the feedback data from students and generates beautiful analytical reports on faculty. This project is used by KMIT with thousands of students to assess their faculty members.

      djangomysql

      2017

  • Games

    • Alster AR

      Alster AR

      A fun AR wrestling game. The game finds a surface on the table and drops the players (balls) on top of it. Your goal is to control the red ball and push the AI-controlled ball out of the table. Whomever falls off the table more number of times in 1 minute loses.

      unityar

      2020

More projects

Blogs

I write articles on Medium

Medium

More about me

  • I am a huge space nerd
  • I sketch really well
  • I like reading books. My books list @ good reads
  • My older blog from undergrad