Nov 2019 – Sep 2023 · 3 yrs 10 mos · Montreal, Quebec, Canada

• Collaborate with clients and prospective clients to design and deliver tailored AI solutions.
• Deliver talks on Reinforcement Learning With Human Feedback (RLHF) and the company’s offerings at conferences.
• Provide strategic and technical direction to ML projects and own the ML roadmap.
• Uphold and refine best development practices.
• Led the design and development of LLM-powered and RLHF-based agents such as Air Traffic Controllers, Battery Management Agents, Fighter Pilot Assistants, Emergency Response Agents, and Forest Fire Detection Agents.
• Built CogmentVerse, a research platform for Human-in-the-loop learning (HILL) and multi-agent Reinforcement Learning (MARL). https://github.com/cogment/cogment-verse
• Integrated Reinforcement Learning algorithms, Deep Learning models, Environments and Technologies into the company's offerings.
• Designed and deployed a variety of innovations such as dimensionality-aware attention in transformers, energy-based parameter-initialization of deep learning models, and modularized training of large neural networks.

Sep 2019 – Oct 2019 · 1 mo · Toronto, Canada Area

• Instructor for machine learning and data analytics courses, which cover a variety of topics such as Python, SQL, Tableau, data preparation, data visualization, operationalisation, machine learning, deep learning, and their applications in Natural Language Processing, Computer Vision, and Recommender Systems.

Nov 2016 – Sep 2019 · 2 yrs 10 mos · Bengaluru, Karnataka, India

• Topic modelling through Variational Mixture of vMF, Spherical X-means, and LDA
• Distributed representation of news articles using Doc2Vec, fastText and Transformer Networks
• Document classification with Graph Convolutional Networks
• Celebrity face generation via Variational Autoencoders and GANs
• Semantic embeddings of photo galleries and news images using ResNet-DocVec

Jan 2016 – Sep 2016 · 8 mos · Bengaluru Area, India

• Classification of digital content consumed by end-users using ResNet and LSTM Networks
• User clustering through Variational Mixture of Gaussians
• Temporal modeling of the users’ buying patterns with ARIMA

Aug 2015 – Nov 2015 · 3 mos · Mumbai Area, India

• Search result design and property recommendations using Probabilistic Graphical Models
• User conversion prediction through CNN-LSTM on real-estate images

May 2014 – Aug 2015 · 1 yr 3 mos · Mumbai Area, India

• Traffic speed prediction through Bayesian Basis Function Regression
• Location and event detection from social media posts and news articles using Probabilistic Models

Feb 2012 – Apr 2014 · 2 yrs 2 mos · Chennai Area, India

• Wind speed prediction over a horizon of three days using an ensemble of Neural Networks
• Optimization of bid strategy in wind energy markets through Dynamic Programming
• Outlier detection for energy consumption patterns at the onshore wind power plants via Probabilistic Graphical Models