Experience

Resolution Games

• Sweden
• Entertainment Providers
• 100 - 200 Employee

• Machine Learning Engineer

  • Sep 2021 - Present

  Broadly applying human-centred ML to creating exciting new player and developer experiences in AR/VR games. • Reinforcement learning and imitation learning research and development applied to shipping vital player-facing game features in upcoming games with on device in-game model inference. • Generative models in Fish Under Our Feet: Developed and shipped real-time on-device animation synthesis system for an AR experience with a fully procedural dancing NPC combining novel dance movement synthesis and movements inspired by player dancing. Created CNN-based generative model (VAE-based) architecture for NPC, extracted motion data from video datasets, and distilled original encoder to use head and two hands inputs from VR/AR headsets. Read more here (https://tinyurl.com/rg-dancing-ai) • Game prototypes exploring AI/ML features: 1) A short VR experience gesture interaction prototype for generating image assets in real-time with a local deployment of Stable Diffusion. 2) A short VR experience using speech commands to control NPCs and GPT4 to generate action plan fragments for NPCs based on user commands, game state, and NPC action sets. • Leading the ML team: strategic direction, project planning, recruitment, mentorship, knowledge sharing, and related responsibilities as team lead. Show less



Microsoft Research

• Think Tanks

• Researcher

  • Aug 2019 - Jul 2021

  Researcher, Deep Reinforcement Learning for Games Group Summary: End to end user + ML research on designer reinforcement and imitation learning by studying workflows + challenges; identifying opportunities for impact; and redefining RL workflows through prototype models + tools. • Studied 17 game AI/ML creators using interviews + thematic analysis to identify 10 challenges and opportunities for researchers in commercial games with Best Paper Award at AIIDE 2020 conference. • Implemented, trained, and deployed InfoGAIL to tweak model w/o retraining by interpolating expert behaviours. • Mentored research intern on adapting agent aesthetic style using preference learning + potential-based reward shaping (PBRS) + automatic reward weighting with a Microsoft Research blog post. • Redesigned workflow, wrote human-in-the-loop RL methods, and made prototype designer tool to improve model robustness from domain distribution shift for product partner game scenario. • AIIDE 2020 best paper award, 2 publications, 1 blog post, 2 prototype tools, 3 invited talks, 1 PhD intern. Show less



Expressive Machinery Lab (formerly Adaptive Digital Media/ADAM Lab)

• Ph.D. Candidate

  • Aug 2017 - Aug 2019

  Ph.D. candidate researching how artificial intelligence (AI) and interactive machine learning (ML) techniques can enable computers to collaborate creatively and improvise with people as well as how to create human-computer improvisational experiences and interactive installations for non-experts.Two examples of my research in this area include the Robot Improv Circus virtual reality (VR) installation (https://expressivemachinery.gatech.edu/projects/robot-improv-circus/) and the LuminAI installation (https://expressivemachinery.gatech.edu/projects/luminai/). LuminAI explored how to interactively learn the knowledge required for performing open-ended movement improvisation with people in real-time focusing on embodied interactions like mimed actions and dance. The Robot Improv Circus extends this work to understand how to use interactively learned knowledge to perform improvisational action selection and decision-making in vastly open-ended problem domains such as farcical object-based improvisation using objects in the agent's environment (i.e. the Props game from improv theatre). The agent attempts to use its knowledge to follow a designer-supplied 'creative arc' over the course of the improvised performance in order to provide the user with a temporally-evolving experience. Show less

• Graduate Research Assistant

  • Jan 2012 - Aug 2019

  Currently researching how computational techniques for evaluating agent creativity can be applied to a co-creative agent for use in the EarSketch music and computing education environment(https://earsketch.gatech.edu/landing/#/).Previously researched a wide selection of topics (reverse chronological order):- Designing/building the hardware for an interactive tabletop museum exhibit called TuneTable where interactors could create music using a tangible interface that encouraged the use of computational thinking concepts.- Computationally performing object substitution and object blending in an intelligent agent in order to pretend that a real-world object is a pretend object and vice versa within the context of human-agent and human-robot toy-based pretend play.- Automatically crowdsourcing knowledge for virtual agents to use in the Three Line Scene improv game for establishing the platform of a scene.- Computationally recognizing and portraying the status of an improvisational virtual character by studying improv theatre actors. Show less



Adaptive Digital Media (ADAM) Lab, Georgia Tech.

• Ph.D. Student

  • Aug 2013 - Jul 2017

  Ph.D. candidate researching how artificial intelligence (AI) and interactive machine learning (ML) techniques can enable computers to collaborate creatively and improvise with people as well as how to create human-computer improvisational experiences and interactive installations for non-experts. Ph.D. candidate researching how artificial intelligence (AI) and interactive machine learning (ML) techniques can enable computers to collaborate creatively and improvise with people as well as how to create human-computer improvisational experiences and interactive installations for non-experts.



Institute for Creative Technologies

• United States
• Research Services
• 1 - 100 Employee

• Visiting Research Assistant

  • May 2016 - Aug 2016

  Researched how to integrate a small set of appraisal models for improving reasoning within the decision cycle of the probabilistic graphical models-based cognitive architecture, Sigma, under the advisement of Dr. Paul Rosenbloom. Researched how to integrate a small set of appraisal models for improving reasoning within the decision cycle of the probabilistic graphical models-based cognitive architecture, Sigma, under the advisement of Dr. Paul Rosenbloom.



Design Intelligence Lab, Georgia Institute of Technology

• Volunteer Researcher

  • Sep 2011 - Dec 2012

  Worked on Self Adaptive Agents Project, designing agents that can adapt themselves to changing rules and game environments. Worked on Self Adaptive Agents Project, designing agents that can adapt themselves to changing rules and game environments.



Citrix R&D India Ltd

• Software Engineer - Intern

  • Jan 2011 - Jul 2011

  Developer working with NetScaler Manageability Team Developer working with NetScaler Manageability Team



Microsoft IDC

• Software Development Engineer - Intern

  • Jun 2010 - Aug 2010

  Dev Intern at MS IDC in the Remote Desktop - Virtualization Team Dev Intern at MS IDC in the Remote Desktop - Virtualization Team



SecureSynergy

• Computer and Network Security
• 1 - 100 Employee

• Intern - Audit team

  • Jun 2009 - Jul 2009

  Learnt about tools, methodologies, frameworks and techniques used by industry professionals in vulnerability assessments and penetration testing Learnt about tools, methodologies, frameworks and techniques used by industry professionals in vulnerability assessments and penetration testing


Malayala Manorama Co. Ltd
• India
• Newspaper Publishing
• 700 & Above Employee
• Intern - IT Systems Team

  • Jun 2008 - Jul 2008