• Leading and mentoring the team of ML Engineers, driving the design, development, and maintenance of Javis' core AI stack - an advanced document parser capable of comprehending and parsing diverse document types with near-100% precision and human-in-the-loop fallbacks.• Pioneering the use of novel multimodal LLM architectures to build the next-gen document parsing stack at Javis.• Envisioned the design and led implementation of a highly elastic architecture capable of scaling to… Show more • Leading and mentoring the team of ML Engineers, driving the design, development, and maintenance of Javis' core AI stack - an advanced document parser capable of comprehending and parsing diverse document types with near-100% precision and human-in-the-loop fallbacks.• Pioneering the use of novel multimodal LLM architectures to build the next-gen document parsing stack at Javis.• Envisioned the design and led implementation of a highly elastic architecture capable of scaling to process 1000s of documents simultaneously, using containerized applications on a serverless platform.• Enabled Javis to expand internationally and enter fresh industries, thereby significantly increasing annual revenue. Show less

- Developed an automated document parser for PDF purchase orders across Major Consumer Goods and Pharmaceutical companies with near-100% parsing accuracy and human-in-the-loop fallbacks.- Architected and implemented a highly scalable architecture for simultaneously processing large volumes of documents using Docker containerised applications on a serverless platform.- Built an FAQ-answering chat interface for educational institutes using PyTorch based siamese-BERT networks. Achieved… Show more - Developed an automated document parser for PDF purchase orders across Major Consumer Goods and Pharmaceutical companies with near-100% parsing accuracy and human-in-the-loop fallbacks.- Architected and implemented a highly scalable architecture for simultaneously processing large volumes of documents using Docker containerised applications on a serverless platform.- Built an FAQ-answering chat interface for educational institutes using PyTorch based siamese-BERT networks. Achieved ~90% recall and ~95% precision in the answered questions. Show less

- Worked on a smart order prediction system to estimate future retailer purchase behaviour.- Developed a clustering system to group retailers with similar characteristics to discover available cross-sell and up-sell opportunities.

Worked on Machine Learning techniques to streamline processes like employee attrition prediction, approval status of generated requests, and sentiment analysis on annual company-wide reviews.

Worked under the bqResearcher programme at Bohr. Conducted research on Quantum algorithms which can be applied to solve classical optimisation problems such as TSP and Graph colouring.

Mentoring second year computer engineering students for Front-End Web development at Unicode.

Worked on the front-end of Unicode's Book Exchange Web App. Used ReactJS for planning components and also basic HTML, CSS, and JavaScript.

Developed Felix Educare's current, live website from scratch, using Webflow.

Was involved in the overall planning and development of an interactive bot on the Telegram messaging platform, which will serve as the primary channel of communication between the company and its clients on a daily basis. The bot is used for data collection and analysis of the user's fitness. The collected data is then logged to Google Sheets for easy access by the fitness coaches appointed to the client.