· Using Deep RL and NLP techniques to focus on the issue of generalisation and Transfer learning in text-based games. · Using PyTorch and state of the art language models...more such as BERT to create better representations to enable transfer learning in these games
Graduate Research Assistant
2018 - Present (almost 2 years)
· Created from scratch, an intelligent agent that plays text-based games from feedback alone. This was done under the supervision of Marc G. Bellemare (Google Brain) and Doina...more Precup (Deepmind, McGill). · Proposed algorithmic improvements for applying Deep Reinforcement Learning techniques to these games. · Worked with PyTorch and used LSTM, word embeddings to create meaningful textual representations. · Submitted this work as a conference submission to AAAI 2020.