Saikrishna Chaitanya Kanala
XXXX@XXXX.XXX | (XXX) XXX-XXXX | https://www.linkedin.com/in/saikrishnachaitanyakanala EDUCATION
University of Southern California, Los Angeles (3.7 gpa)
Aug 2018 - May 2020
Master of Science in Electrical Engineering (Specializing in Machine Learning)
Courses: Linear algebra, Probability, Pattern Recognition, ML from signals, Deep Learning, Applied ML for games
University of Pune, India Aug 2012 – May 2016
Bachelor of Engineering in Electronics and Telecommunications
INDUSTRY EXPERIENCE
Mobworx | Machine Learning Intern | Los Angeles May 2019 – Nov 2019
• Produced a deep learning GAN model for super resolution of videos, converted it to coreML model
• Reduced the size of the videos to at least 60%, over the bandwidth and implemented neural style transfer
including age and gender classification
• Trained and tested algorithms like VDSR, SRGAN, EHGAN, FSRCNN, benchmarked al the models for speed
(frames per second) and quality (frame quality) using fritz.ai CLI
Infosys Ltd | Test Engineer | Pune Aug 2016 – Jun 2018
• Conducted data migration of client’s Financial Services databases employing SQL, Informatica and Hp-QC
• Generated virtual services like web services, rest services, IBM message queues using CA Dev Test LISA 10.0
tool for testing using request and response xmls RESEARCH Jan 2020 - Present
• Research student at USC Hardware Acceleration Lab, working on AutoML search using Bayesian optimization
• Devising a method to identify high accuracy producing models (greater than 90%) within shortest time
• The optimization includes Neural Architecture Search as well as hyperparameter tuning with results
comparable to Auto-Keras (Paper currently submitted to ECML-PKDD 2020)
SKILLS Languages and Machine learning frameworks:
Agile and version control Tools:
Python, MATLAB, C++, SQL Jira, Confluence, Bitbucket, Git
Keras, Tensorflow, Scikit-Learn, Pytorch, Pandas, OpenCV PROJECTS
Neural Aim bot and Reinforcement learning in fps game (CS + Doom) Aug 2019 – Dec 2019
• Designing an algorithm for deep object recognition using SSD for detecting different players in the game
• Triggering auto fire when the opponent is detected, and crosshair fal s on the person
• Taught the bot from Doom game to learn how to navigate till the end of the map, avoid shooting and take on
maximum enemies at once
Emotion Detection from facial expressions using CNN Feb 2019 – Apr 2019
• Trained a CNN on the AWS cloud server using the given data set of images containing faces with emotions
• Determined the face and classified the expression in the given image. Used CNN to build a 9-layer dense
neural network and used max-pooling to down sample
Explorations in Deep Reinforcement Learning Apr 2019 – May 2019
• Built three different reinforcement learning models: Random Network Distil ation, Distributional RL, Noisy nets,
and apply these models in functional Atari 2600 games like ‘Montezuma’s Revenge’ and ‘Solaris’
• Selected the best model amongst the three, based on complexity of the network, number of trainable
parameters and the most efficient policy selection, to extend this model in live application for the localization
of the quadcopters and self-driving (https://www.youtube.com/watch?v=f8pR1e7SszA)
Predicting the winner of tennis matches in ATP Feb 2019 – May 2019
• Estimated the likelihood of “wins” in tennis matches of men’s ATP tour
• Implemented data preprocessing techniques like normalization, one hot encoding, class imbalance along with
feature extraction and dimensionality reduction including a high priority user-defined feature ‘log of wins’
• Compared classifiers like Logistic Regression (0.7), SVM (0.51) and Random Forest (0.64) on accuracies
PUBLICATIONS
• Virtual Keyboard; (IJAR) ISSN 2(XXX) XXX-XXXX Volume 4 Issue X2 Dec 2016
• Sixth Sense Technology: The Way Forward; (IJRASET) ISSN: 2(XXX) XXX-XXXX; Volume XXXXXX 2017