Amal Joe R S

Autumn 2024 - Teaching Assistant for CS101: Computer Programming and Utilization Spring 2025 - Teaching Assistant for CS104: Software Systems Autumn 2025 - Teaching Assistant for CS419: Introduction to Machine Learning

Integrated agentic AI into an existing financial report creation workflow to automate iXBRL tagging and fine-tuned custom models for in-house use cases. The entire pipeline was built from scratch and there were no reference articles or datasets in place. Web scraping was done to prepare a dataset of 10M+ tokens. It was then augmented with synthetic data and used for finetuning. RL was used to further increase the tagging accuracy. XBRL tagging is a classification problem but with enormous number of classes than a standard problem. The final model achieved 80% tagging accuracy on a held out validation set with 12,000+ different tags in corpus.

- Designed a dynamic data sampling strategy that optimizes multi-task LLM training by measuring a model's learning potential. This approach enhances training efficiency by preventing resource waste on tasks the model has already mastered. It beats state of the art static and dynamic sampling methods by a considerable margin (5-10% improvement in down stream task performance) - Integrated this dynamic data loader with IBM’s internal training stack and conducted trainings on real world use cases of IBM. We observed considerable improvements compared to state of the art static and dynamic sampling methods.

- Implemented code-splitting using dynamic import in React application and optimized chunking with Rollup's manualChunks configuration, resulting in a 40-50% improvement in application load times - Used module federation to split a monolithic React application into different micro frontend applications enhancing better collaboration among cross-functional teams and reducing development time by 20-30%

- Elected as class representative in pre-final year - Elected as department representative in final year - Conducted industrial visit for CSE department - General Convenor for department technical fest, HASH 2022 - Technical Head for college cultural fest, CROSSROADS 2023 - Led the technical team of various organizing committees for different events and celebrations at college (Onam Celebration, Christmas Celebration, Ethnic Day etc..)

Represented the Microsoft Learn program and organized events to promote technology education and engagement.

Conducted seminars on Intel's cutting edge technologies in AI domain, reaching an audience of 200+ students.

- Participated in the Android Study Jams 2021 conducted by Google Developers India as a Campus Facilitator - Led the technical team to conduct a 3 day design workshop, Uxtopia, in partnership with Figma. Handled the media works of live streaming the event to YouTube including real-time processing of green screen footage. - Led a three-day Flutter workshop for juniors, enhancing their app development skills through hands-on sessions.

- Developing the primary mobile application of the company - Integrating detailed user analytics into the app - Improving crash analytics and improving the performance of the app - End to end management of the app on playstore

- SGPAs - S1: 9.54, S2: 9.63, S3: 9.90 - Specialising in AI/ML coursework (Deep Learning for NLP, Foundations of ML, Foundations of Intelligent and Learning Agents, "Speech, NLP and the Web" etc...) - R&D-1 on "Explainability in LLMs" - R&D-2 on "LLM fine-tuning and customisation" - Seminar on "Data selection during fine-tuning using sub-modular optimisation" - MTP (Master Thesis Project) on "Online data mixing during fine-tuning"

- Graduated First Class with Distinction (CGPA 9.31) - Proficiency Award (1st Rank) recipient in the Computer Science and Engineering department - Nominated for the Baselian Award (all-rounder) on graduation and recipient of the Best Performing Student Award - Relevant Courses: Artificial Intelligence, Soft Computing, Programming in Python, Image Processing, Computer Graphics, Data Compression, Distributed Computing, Seminar on Style Generative Adversarial Network

Completed semester 1 of BTech at SRM university and later took an inter university transfer to APJ Abdul Kalam Technological University.