About Me

Hi, I'm Sharan Shyamsundar. I'm a Data Scientist and AI Engineer passionate about transforming data into innovative AI-driven solutions. I thrive on solving complex problems with AI and data. Over the past 5 years, I've built a track record of delivering practical, impactful solutions in AI, Machine Learning, and Analytics.

LLM Fine-Tuning & RAG Systems: Designed and optimized LLM pipelines for domain-specific tasks, ensuring efficiency, relevance, and scalability.
Financial Analytics & Classical ML: Developed predictive models and risk assessment frameworks, enhancing decision-making in financial sectors.
Classical Machine Learning & Data Engineering: Engineered intuitive tools and interactive dashboards that streamlined decision-making processes, improved data accuracy, and optimized operations across diverse industries.

What sets me apart? I don’t just build models—I build solutions that work in the real world. Whether it's fine-tuning an LLM for better reasoning, optimizing a retrieval system for accuracy, or engineering scalable AI workflows, I bring a mix of deep technical expertise and a sharp problem-solving mindset.

I'm open to AI Engineer/Data Science opportunities—feel free to connect if you'd like to collaborate!

Projects

PEFT methods applied to Quantized LLMs

Explored the integration of Quantization and Parameter-Efficient Fine-Tuning (PEFT) methods to optimize LLM finetuning. While Quantization reduces memory and computational demands by compressing model parameters, PEFT minimizes the number of parameters updated during fine-tuning. Experimental results demonstrate that certain Quantization and PEFT pairings enable substantial improvements in memory efficiency and computational performance, paving the way for practical deployment of LLMs in resource-constrained environments.

Implementation Report

Report

Foodle - Personal AI Cooking Assistant

Foodle is an innovative cooking assistant that leverages Automatic Speech Recognition (ASR) and LLMs to enhance the culinary experience. Designed for amateur chefs, it transcribes cooking videos into text, generates recipes, and offers hands-free voice interactions. Users can ask questions about recipes, cooking techniques, and ingredient substitutions, enabling real-time assistance without the need to rewatch videos.

Implementation

Multi Agent Article Writer

This App is an AI-powered multi-agent framework designed to automate the content creation process, replicating the workflow of a professional writing team. Using CrewAI, the system assigns specialized agents for planning, writing, and editing, ensuring high-quality, well-structured articles. The framework optimizes collaboration between agents to refine content iteratively, enhancing coherence and readability. It can be customized for different writing styles and topics, making it a versatile tool for bloggers, marketers, and businesses.

Implementation Hugging Face Spaces

HF Space

ZENIX - Zero Effort Natural language to Integrated XML

ZENIX was developed as part of the Hackathon@WEF25. It is a multi-agent, robust AI solution that seamlessly translates natural language into precise SQL queries while generating the necessary XML business objects. To enhance the user experience on the JiVs IMP platform, we integrated a feature that intelligently generates navigation links to desired tables and views.

Implementation

Pitch Perfect

This was developed as a personal project of mine. Create personalized, job-specific cover letters in minutes with Pitch Perfect. Using intelligent prompts and customizable templates, it helps you highlight your experience and skills for every application. Whether you're a seasoned professional or just starting your career, Pitch Perfect saves time and boosts your chances of landing interviews—making every application stand out. Do try it out on HF Spaces.

Implementation Hugging Face Spaces

HF Space

Multi Agent Project Planner

Multi-Agent Project Planner is an AI-powered app built with CrewAI. Just enter your project's details—like name, industry, objectives, specific requirements, and team members—and it automatically creates a clear, step-by-step project plan. Tasks and milestones are broken down and intelligently assigned to each team member, helping your team stay organized, efficient, and focused on successful project delivery. Do try it out on HF Spaces.

Implementation Hugging Face Spaces

HF Space

Articula (Informatics Cup 2024)

ARTICULA is a solution for detecting and spoofing AI-generated text and images to combat misinformation and copyright infringement. It uses ensemble models that combined low-resource, domain-agnostic detection techniques for text and images, achieving enhanced accuracy in identifying AI-generated content. Implemented iterative spoofing attacks leveraging model feedback to introduce minimal perturbations, effectively bypassing detection mechanisms.

Implementation

Multi Agent Event Planner

This App is an AI-powered multi-agent framework designed to automate the entire event planning process with intelligent multi-agent collaboration. By gathering key details like the event topic, location, date, budget, and expected participants, it efficiently searches for venues, organizes logistics, and generates a comprehensive marketing report. Designed for seamless execution, this framework simplifies event management, making planning faster and more data-driven.

Implementation

Mobility Analytics

A data-driven solution designed to analyze surge prices for a Cab Company and optimize urban mobility. Analyze the acquired data of surge pricing types assigned to consumers using various machine learning approaches and constructed Random Forest, SVC and KNN models that can help in pro-actively predicting surge pricing types utilizing factors such as trip distance, cab type, customer lifestyle details, and much more.

Implementation

Pitch Perfect V2 (In Progress)

This app takes Pitch Perfect to the next level by leveraging a multi-agent framework powered by CrewAI. It goes beyond simply creating cover letters by also offering personalized resume suggestions and generating tailored sample interview questions to help users practice. With this expanded functionality, the app provides a comprehensive solution for job seekers, supporting them through multiple stages of the application process.

Implementation (Will be updated soon)

About Me

Projects

PEFT methods applied to Quantized LLMs

Foodle - Personal AI Cooking Assistant

Multi Agent Article Writer

ZENIX - Zero Effort Natural language to Integrated XML

Pitch Perfect

Multi Agent Project Planner

Contact