skip to content
Shreyas Devaraj

Projects

Here are some of my key projects that demonstrate my expertise in AI/ML, Software Engineering, and Full-Stack Development.

Eventflow.ai – Agentic Application

An innovative agentic assistant that integrates multiple LLMs (ChatGPT, Llama, Gemini) to automate task generation through multi-agent reasoning. Winner of $3500 prize at VenturePack Challenge.

Key Highlights:

  • Multi-agent reasoning with LLMs
  • Full-stack architecture with FastAPI and TypeScript
  • Interactive chat with dynamic event understanding
  • Won $3500 at VenturePack Challenge

Tech Stack:

LangGraph Langchain Python FastAPI TypeScript CopilotKit

d-shop Assistant

A Generative AI assistant for SAP d-shop internal wiki and support tooling using LLMs and RAG techniques. Integrated with Teams, GitHub, and Outlook to streamline workflows.

Key Highlights:

  • Context-based responses using RAG techniques
  • Integration with Teams, GitHub, and Outlook APIs
  • Reduced manual query handling by 60%
  • Prompt engineering and task orchestration

Tech Stack:

Python GPT Langchain Graph API

Question Answering System using Knowledge Graphs

An NLP model incorporating Knowledge Graphs to answer complex multi-hop questions. Achieved 80.13% accuracy, outperforming BERT on SQuAD and Multi-RC datasets. Published in IEEE Xplore.

Key Highlights:

  • 80.13% accuracy on multi-hop questions
  • Outperformed BERT on SQuAD and Multi-RC datasets
  • Published research paper in IEEE Xplore
  • Entity pair extraction and POS tagging

Tech Stack:

Python spaCy NLTK networkX

BioInsight - Biomanufacturing Analytics Platform

End-to-end analytics platform for BTEC to analyze biolector data from 1mL and 30L scale experiments. Built ML pipelines for KPI prediction and time series forecasting.

Key Highlights:

  • ML pipelines for KPI prediction and forecasting
  • LSTM for time series forecasting
  • Isolation Forest for outlier detection
  • Interactive Streamlit visualizations

Tech Stack:

Python Scikit-Learn TensorFlow Matplotlib Streamlit

Emotional Virtual Assistant (EVA)

A sophisticated chatbot designed to provide mental health support and answer COVID-19 related FAQs. Built using advanced NLP techniques to understand user emotions and provide appropriate responses.

Key Highlights:

  • Emotion detection and response generation
  • COVID-19 FAQ integration
  • Mental health support capabilities

Tech Stack:

Python RASA NLP

StoryHouse - E-commerce Platform

A comprehensive e-commerce website for books built using the MERN stack, featuring modern UI components developed with StencilJS for enhanced user experience.

Key Highlights:

  • Full-stack MERN implementation
  • StencilJS component integration
  • Modern e-commerce features
  • Responsive design

Tech Stack:

MongoDB Express.js React Node.js StencilJS