Through my work with the blind chess community, I strive to make the wonderful game of chess more accessible to visually impaired players.
Indian National Coach – Junior Blind Chess Team
Jun–Oct 2024
As the National Coach, I have trained my students for global tournaments such as the Asian Paralympics and World Juniors.
I had the privilege of mentoring our team and helping them prepare for the 12th IBCA World Junior Chess Championship for the Blind held in Bengaluru in Oct 2024. My work focused on developing their tactical chess and strategic skills and boosting their confidence, ensuring they perform at their highest potential at the international level.
This Chessbase India article has more about my work.
HelpChess Foundation – Asian Paralympics
Jun–Nov 2023
I have worked closely with Sanskruti More to help her prepare for the 2023 Asian Paralympic Games, where she won a Bronze medal! She was mentioned by the Indian Prime Minister Shri Narendra Modi for her achievements. (Govt. of India – PIB Link)
Invision Chess Academy
July–Oct 2024
I founded the Invision Chess Academy to connect visually impaired players with passionate mentors from all around the world, providing personalized coaching and support to help them develop their skills. The academy fosters an inclusive environment where players of all abilities can thrive, empowering them to compete at higher levels while promoting accessibility in chess.
If you are a visually impaired player interested in learning in a new, engaging way or an enthusiastic sighted player who would like to mentor or play practice with blind players, connect with us!
Chesscript.AI
Ongoing (7 months)
I am developing Chesscript.AI, an app that uses CNNs (Convolutional Neural Networks) to convert chess book PDFs into the accessible PGN format. This innovative tool aims to provide equal training opportunities by giving visually impaired players access to the same resources as sighted players.
My project was more than just community service: it was an immense learning experience in artificial intelligence and machine learning.
I aimed to convert PDF chess books, with all their diagrams and annotations, to a text-based PGN format. The development process involved two components: chessboard (image) recognition and game (text) recognition.
For chessboard recognition, I first identified the location of the chessboard on the PDF using the colour disparity between the white page and the edge of the board. I then split the chessboard into its 64 individual tiles. Next, to create an intelligent model that could understand the piece on each tile, I manually created a database and used TensorFlow’s high-level API, Keras, to create a convolutional neural network (CNN). This ML model, with enough training data, could correctly identify the chesspieces!
Text recognition was tougher. I first extracted the raw text into a markdown file using a Python library. But in order to meaningfully understand this markdown, I had to incorporate a large language model pipeline. I used the Gemini API to access Google AI studio, where I listed examples to train the machine learning model to output PGN given markdown.
My application is currently being beta-tested by my eager visually impaired students. We will soon be ready to launch it officially!