Hi, I'm Avishek Biswas
Senior ML Scientist
MS, Clemson University.
First and foremost, I'm a Machine Learning Guy
I have undertaken projects in many different disciplines of ML - Computer Vision, NLP, Reinforcement Learning, Network Science, Robotics, Biomedical Data Science, Autonomous Driving, etc. Some of these projects have been at an industrial scale, some were research projects that led to publications and poster presentations, some were for my Youtube channel, some I did to ace an exam, and some were just crazy adventures I took on a random Saturday afternoon.
Awards and Achievements
Awarded the "Best Paper Award, 1st position" in ACM Siggraph Conference on MIG 2021 for the paper Motor Babble: Morphology-Driven Coordinated Control of Articulated Characters
Awarded the "Outstanding Master's Student in Computer Science" by Clemson University School of Computing on April 2020.
Awarded the "Best Poster Award" in ACM Siggraph Conference on Motion, Interaction and Games (MIG) 2020.
Background
I am working as a Data Scientist at Quiq, one of the fastest-rising startups in the world of customer-business chat services. At Quiq, we spend hours developing and deploying cutting-edge Machine Learning and AI techniques to enhance the quality and efficiency of conversations between end-users and live agents. My role requires a ton of NLP and Deep Learning research, and also train models with production deployment and scalability in mind, and striving for quiq releases and immediate product impact.
Prior to working at Quiq, I completed my Master of Science (MS) in Computer Science from Clemson University in May 2021. At Clemson, I worked for three out of my four semesters as a Graduate Research Assistant (GRA) and Graduate Teaching Assistant (GTA). My research experiences at Clemson cover a variety of fields: Artificial Intelligence, Machine Learning, Character Animation, and Network Science.
My Master's Thesis at Clemson involved using Deep Reinforcement Learning to train physics-based character controllers. The thesis covers two research projects - "Learning Coordinated Locomotion from Latent Coactivation Space" and "Training Physics-based Characters to Dance to Music". The former project, titled Motor Babble: Morphology-Driven Coordinated Control of Articulated Characters, recently won the best paper award at MIG (Motion In Games) 2021. The latter project won the Peer-Voted Best Poster Award at MIG 2020.
Industrial Experience
Senior Machine Learning Scientist - Advanced Conversational Techniques, QUIQ INC (May 2021 - May 2024)
This was my first industrial stint as a Data Scientist! I am the first Data Scientist hired at Quiq, which not only put me in a position of responsibility but also accelerated my growth in the field. Being an active part of building end-to-end Machine Learning frameworks, from data cleaning to model training to deployment, has given me a wholesome view of Machine Learning and has been an incredible learning experience on top of my time as a research assistant in grad school.
Key Achievements:
Develop efficient autoregressive text generation models capable of running in browsers at high latency enhancing real-time user interaction by autocompleting agent messages as they type. Our algorithms were released to some of Quiq’s biggest clients contributing majorly towards Quiq’s revenue and major contract extensions.
Train Transformers assisted embedding space to augment conversation states during live chat sessions between customers and businesses, which drive smarter in-app decision-making and live-agent productivity.
Create complex Agent Based Prompting Systems with Large Language Models.
Primary author of non-trivial deep learning model architectures customized to conversational data sets, developing with production deployments and scalability in mind – resulting in quick releases and immediate product impact
Senior Software Developer - Capgemini (2016 - 2019)
I worked with Capgemini India as a Senior Software Developer from Sep-2016 to Mar-2019. I was part of a Oracle Applications DBA (Database Adminstrator) Team. My professional experience working as an Oracle DBA in Capgemini has given me invaluable technical insights into Database Systems, Automation, and Linux. Equally importantly, it has enriched my leadership qualities, team spirit, communication skills, and my ability to deliver my best in critical situations!
Key Achievements:
My team was awarded the Team Award of Excellence for the “Wave 2” Oracle EBS Migration project for Norway Posten
Automated several modules of the project through shell scripts and reduced manual workload of our team. Here is one such example script to automate the transfer of Oracle Concurrent Request Log/Out files from Production to Dev servers: Github Link.
Mentored and trained team of three in Oracle E-Business Suite (EBS) and Linux Environment
Graduate Research Experience
Dec 2019-May 2021: Working with Dr. Victor Zordan, Dr. Ioannis Karamouzas, and Dr. Avinash Ranganath in the intersection of Physics-based Character Control and Deep Reinforcement Learning.
Please visit the Projects page for further details!
Teaching Assistant Experience
Spring 2020: CPSC 4050/6050 - Computer Graphics, by Dr. Daljit Singh Dhillon.
Fall 2020: CPSC 4420/6420 - Artificial Intelligence, by Dr. Ioannis Karamouzas
Spring 2021: CPSC 8810 - Deep Reinforcement Learning, by Dr. Ioannis Karamouzas
Courses
Relevant Courses (Clemson University):
Fall 2019:
CPSC 6420: Artifiicial Intelligence
CPSC 6050: Computer Graphics
CPSC 8040: Data VisualizationSpring 2020:
CPSC 8400: Design and Analysis of Algorithms
CPSC 8810: Motion Planning
CPSC 8430: Deep LearningFall 2020
CPSC 8420: Advanced Machine Learning
CPSC 8480: Network Science
GPA: 4.0/4.0
Extra Curricular Activities
I used to compete in over the board Chess tournaments until couple of years back. In my free-time, I still like to play some Blitz on Chess.com and make Youtube Videos on Machine Learning.