CV
Download:
- CV – last updated 28th Sept 2023
Employment
- AI Scientist III at Electronic Arts (Aug 2022 - Present)
- Researching, designing, and prototyping novel tools for automated player play-style analysis.
- Developed mapping-based unsupervised encoding and segmentation for automated player behavior grouping and analysis.
- Patent application submitted.
- Researching techniques for more generalized gameplay A.I.
- Developed adaptive RL-technique for training more generalized agents.
- Award winner in ‘Tools and Platform’ category division at internal WhatIf technical fair.
- Researching, designing, and prototyping novel tools for automated player play-style analysis.
Education
PhD in Computer Science, Boston University (Aug 2017 - Aug 2022)
- MSE in Robotics, University of Pennsylvania (2016)
- GPA 3.75
- MEng in Mechatronic Engineering, The University of Nottingham (2014)
- High Achievers Scholarship (2010 - 2014)
- Dean’s List of Top Performing Students in the Department of Electrical and Electronic Engineering
Research experience
- Research Fellow - Kate Saenko’s research group and IVC at Boston University (Aug 2017 - present)
- Supervisor: Prof. Kate Saenko
- Generalization and Domain Adaptiation in RL - Investigating techniques to bridge the domain gap when applying control policies learned on one domain to another. Developed automated bandit-based learning scheme for developing Reinforcement Learning control policies that are robust to perturbations in actor environments (paper submission currently under review).
- Neuroflight - Neural Network based Flight Controller for High Performance Racing Drones - Developing novel network architectures and training regimes to improve the accuracy and stability of control output in deploying neural-flight controller models trained in simulation to real drone hardware. Work done in collaboration with Renato Mancuso.
- Reinforcement Learning Aided Design (RAD) - Developing a Reinforcement Learning based pipeline to automatically adjust designs of material composites to satisfy specific physical properties. Work being done in collaboration with Emily Whiting, Keith Brown, Elise Morgan and Wojciech Matusik.
- Neuroflight - Developing neural network based flight control for high-performance quadrotors in collaboration with Bassel Mabsout and Prof. Renato Mancuso
- Supervisor: Prof. Kate Saenko
- AI Intern - Electronic Arts [EA Sports] (September - December 2021)
- Developed novel framework for blending reinforcement and imitation learning for play-styles.
- AI Intern - Electronic Arts [Digital Platform Data & AI] (May - August 2021)
- Developed Reinforcement and Imitation Learning framework to make A.I. agents in games learn to imitate demonstrations or fall back on reinforcement learning when demonstrations are not available.
- AI Intern - Electronic Arts [Digital Platform Data & AI] (May - August 2020)
- Developed Reinforcement Learning tools to make A.I. agents in games learn and play with different behavior styles.
- Developed novel actor-critic training architecture to enable a single actor network to successfully learn multiple play styles, thus significantly reducing runtime cost of using learned controllers.
- Research Assistant - GRASP Lab at the University of Pennsylvania (June 2016 - May 2017)
Supervisor: Prof. Kostas Daniilidis
- Product ID - Worked on developing a framework for efficient autonomous multi-class product labeling and localization in natural image of shelves in stores
- Autonomous flight guided by event-based camera (Group effort) – Vision-based flight control with an event-based camera. Specific responsibilities: State estimation, controller design, hardware construction & management.
Supervising Professor: Prof. Mark Yim
Subgroup: ModLab
Designed a novel, low cost, 0-DoF end-effector and manipulation scheme. (Currently being prepped for peer-review)
- Student Researcher - GRASP Lab at the University of Pennsylvania (June 2015 - May 2016)
- Master’s Thesis (see below)
Other Work:
Supervisor: Prof. Jianbo Shi
- Helped develop and implement real-time image-based camera localization in known 3D spaces (Using feature matching and SfM)
- Investigated application of deep-learning in characterizing the camera motion between sequential images (with Dr. Hyun Soo Park)
- Developed custom object tracking method and software to interface with existing VICON hardware infrastructure and provide real-time pose tracking of large numbers of agents
Dean’s Undergraduate Research Program - The University of Nottingham (June - Aug 2013)
Participant in a selective research and development initiative
Supervisor: Dr. Belle Ooi
- Developed prototype electronics and Graphical User Interface (GUI) for the control of high-performance micro-valves
- Researched and Experimented with potential solutions to low-cost printable circuits and antennae (using standard office printers)
Publications
Mysore, S., Cheng, G., Zhao, Y., Saenko, K., Wu, M. (2022). “Multi-Critic Actor Learning: Teaching RL Policies to Act with Style”, International Conference on Learning Representations, 2022
Gongora, A.E., Mysore, S., Li, B., Shou, W., Matusik, W., Morgan, E.F., Brown, K.A., & Whiting, E., “Designing Composites with Target Effective Youngs Modulus using Reinforcement Learning”, ACM Symposium on Computational Fabrication, 2021
Mysore, S., Mabsout, B., Saenko, K., & Mancuso, R. (2020). “How to Train your Quadrotor: A Framework for Consistently Smooth and Responsive Flight Control via Reinforcement Learning,” ACM Trans. Cyber-Phys. Syst. 5, 4, Article 36 (October 2021), 24 pages. DOI:https://doi.org/10.1145/3466618
Mysore, S., Mabsout, B., Mancuso, R., & Saenko, K. (2021). “Honey, I Shrunk The Actor: A Case Study on Preserving Performance with Smaller Actors in Actor-Critic RL,” IEEE Conference on Games 2021, Virtual Event
Mysore, S., Mabsout, B., Mancuso, R., & Saenko, K. (2021). “Regularizing Action Policies for Smooth Control with Reinforcement Learning,” IEEE International Conference on Robotics and Automation 2021, Xian, China.
- Preprints
Mysore, S., Cheng, G., Zinno, F., Zhao, Y., Saenko, K. (2023). “Split-Critic Imitation Learning for Balancing Conflicting Imitation and Reinforcement Learning Objectives”, Preprint, 2023
Mabsout, B., Roozkhosh, S., Mysore, S., Saenko, K., Mancuso, R. (2023). “The SwaNNFlight System: On-the-Fly Sim-to-Real Adaptation via Anchored Learning”, Preprint, 2023
Mysore, S., Platt, R., Saenko, K. (2019). “Reward-guided Curriculum for Robust Reinforcement Learning.”
- Workshops
- Mysore, S., Platt, R., Saenko, K. (2019). “Reward-guided Curriculum for Learning Robust Action Policies”, Workshop on Multi-Task and Lifelong Reinforcement Learning at ICML 2019
Technical Skills
- Programming:
- Computer Aided Design:
- Electronic Design:
- Circuit Design, PCB design and population
- Machining and Fabrication:
- Trained in: Turning, Milling, Drilling, Tapping, Threading, and Welding
- Additional: Laser cutting, 3D printing
Technical Reports
- University of Pennsylvania
Master’s Thesis - logVLAD: A Novel Pipeline for Image Retrieval
Supervisor: Prof. Kostas Daniilidis
Developed a new variation on the Vector of Locally Aggregated Descriptors (VLAD) Pipeline, using a logarithmically scaled encoding, that presented with superior precision over (then) state-of-the-art and better response to feature burstiness, with good energy distribution
- The University of Nottingham
4th-Year Group Design Project
Supervisor: Dr. Kevin Lee
Specific area of focus: Object Detection and Recognition for robot guidance. Utilized SIFT-based marker recognition to recognize target objects and estimate the distance to the object
3rd-Year Undergraduate Thesis
Supervisor: Prof. Haider Abbas Almurib
Utilized shape and color context to recognize and estimating distances to known objects from monocular images
Selected Coursework
- University of Pennsylvania
Reconstructing 3D Scenes from Image Sequences Applied concepts of Structure from Motion (SfM) and epipolar geometry to build sparse 3D scene reconstructions from a sequence of images.
Stereo Visual Odometry Implemented and compared different approaches to stereo visual odometry against the KITTI vision benchmark suite. Implementations computed odometry by solving the 3D-3D affine Procrustes problem, by solving the 3D-2D Perspective-n-Point (PnP) problem, and by using optical flow.
Learning Path-planning Extracted feature-maps from a satellite-view map using Gaussian Mixture Models (GMMs) trained to recognize colors, and used them to build a cost-map over which Dijkstra’s and A* algorithms were applied to determine the best traversable path between arbitrary start and goal set on the map.
Face Replacement in Images Developed a program that attempts to seamlessly replace faces detected in an image with some other face, with skin-tone blending and masking to account for face rotations.
Gesture Recognition Applied Hidden Markov Models towards learning to recognize hand gestures utilizing inputs from an Inertial Measurement Unit (IMU) attached to the arm.
Autonomous Quad-rotor Flight Implemented programs to facilitate autonomous path-planning and flight control in real-time on research- and consumer-quad-rotors.
Light-Painting with a Robot Arm Modeled DH-parameters and Inverse Kinematics of a robot arm to plan and execute the ‘painting’ of a picture with an LED mounted at the arm’s tool-tip, captured on a long-exposure image.
Machine Learning applied to Pricing Estimation Utilized several machine learning techniques attempts to estimate a price range for real-estate listings using bags of words provided to describe each listing.
Teaching
- Boston University
Teaching Assistant CS 542 Machine Learning, Spring 2022
Grader CS 581 Computational Fabrication, Spring 2022
Guest Lecturer CS 542 Machine Learning, Fall 2021
Grader CS 440 Artificial Intelligence, Fall 2020
Grader CS 480/680 Introduction to Computer Graphics, Fall 2019, Fall 2020
Grader CS 542 Machine Learning, Fall 2018, Spring 2020
- University of Pennsylvania
- Teaching Assistant
CIS 581 Computer Vision and Computational Photography, Fall 2015
- Teaching Assistant
Service
Reviewer for the IEEE Robotics and Automation Letters (RAL) 2023
Reviewer for the IEEE International Conference on Robot Systems (IROS) 2022, 2023
Reviewer for the IEEE International Conference on Robotics and Automation (ICRA) 2022
Reviewer for the Conference on Neural Information Processing Systems (NeurIPS) 2021, 2022
Reviewer for the Conference on Robot Learning (CoRL) 2021, 2022, 2023
Reviewer for the IEEE Conference on Games 2021, 2022, 2023
- Coordinator for AI Research (AIR) Seminar Series at BU (Aug 2018 – Aug 2020)
- Manage hosting and planning duties for the AIR group’s seminar series
- Coordinator for Image and Video Computing (IVC) group meetings and website at BU (Aug 2018 – Aug 2021)
- Manage planning duties for IVC’s weekly meetings and also manage the group’s website
- Technical coordinator for AI4ALL Summer Program (May 2019 – Aug 2019)
- Trained AI4ALL Undergraduate interns at Boston University to develop and execute code for Machine Learning
- Reviewer for the Machine Learning Journal (Jun 2019)
Industrial Training
Viyas Innovative Technologies Pvt. Ltd. & Staysee Healthcare Products (P) Ltd.
Mysore, India (June - Aug 2012)
Obtained a comprehensive understanding of the processes involved in design, manufacture, assembly and quality control of electronic devices in medium-scale industry
Assisted process engineers in day-to-day activities
Language Skills
English (1st Language)
German (Certified A2, Studied B1)
Kannada (Basic conversational ability)