Anurag Shenoy

I’m a Software Engineer based in the US, passionate about AI-driven solutions.

Currently open to work, I’m looking for opportunities to make an impact.

Scroll down to see how I bring ideas to life

Project 01

MedQuAD RAG Bot — A Retrieval Augmented Generation Chatbot

A Chatbot utilizing the MedQuAD Dataset to provide accurate medical information and answers using technologies such as Retrieval Augmented Generation (RAG) and Large Language Models (LLMs).

Tech Stack: GPT 4o-mini LLM, Qdrant Vector DB, LangChain framework, FastAPI, & React.

Deployment: AWS Lambda, Docker, AWS ECR, AWS API Gateway, Netlify.

Key Features:
  • Retrieval using Qdrant’s Vector DB, with the MedQuAD dataset as it’s high-quality, and accurate knowledge base.
  • Prompt Engineering for providing context and constraining output length and scope.
  • Deployment on Serverless compute for cost-effectiveness and high-throughput.

Links: MedQuAD RAG Bot | GitHub

Demonstration screenshot of HinDict the Hindi to English dictionary project using the word Namaste as input

Project 02

HinDict — Hindi to English Dictionary

A bilingual dictionary that lets users type Hindi words phonetically in English letters to find the best English equivalents.

Tech Stack: OpenAI GPT 4o-mini, React, Nextjs, Python, Flask, Ollama, and Prompt Engineering.

Key Features:
  • Leverages advanced AI models to parse transliterated Hindi text.
  • Provides definitions, pronunciations, and contextual examples.
  • Aims to simplify language learning for non-native Hindi speakers.

Links: HinDict | GitHub

Demonstration screenshot of HinDict the Hindi to English dictionary project using the word Namaste as input

Project 03

Debate AI — Agentic Research-Paper Debate Tool

An autonomous debate platform that pits two LLM-based AI agents against each other on user-selected topics, moderated by a third AI agent.

Tech Stack: RAG, LangChain, LLM Inference, Ollama, Weaviate Vector DB, React, and Prompt Engineering.

Key Features:
  • Uses retrieval-augmented generation (RAG) and vector search with Weaviate to provide contextually rich arguments.
  • Demonstrates how AI agents can interact to refine complex ideas.
  • Serves as a unique learning tool where users can observe AI-driven debates on cutting-edge research.

Links: Debate AI

Demonstration screenshot of Debate AI project, with two AI agents debating the topic "Should transformer model be used as the primary architecture for all NLP models?".

Project 04

Aggre-Gator — Hackathon-Winning News Aggregator

A web platform that scrapes news from sources like CNN and Fox, allowing users to view both sides of a story on a single dashboard.

Tech Stack: ElasticSearch, Docker, MongoDB, Python, Flask, Celery.

Key Features:
  • Centralizes articles from multiple outlets, promoting balanced news consumption.
  • Provides search and filtering with Elasticsearch for near-instant results.
  • Automates data collection and indexing via Docker + Celery for reliability.

Links: Aggre-Gator GitHub

Screenshot of Aggre-Gator landing page, with news articles aggregated from CNN and Fox.

Project 05

Guessir - Word Guessing Game with a Twist

A 6-letter word-guessing game featuring a dynamic “Hot & Cold” mechanic to guide players on letter positions.

Tech Stack: React, Nextjs, Python, FastAPI, Postgres

Key Features:
  • Uses color gradients to indicate how close each guessed letter is to its correct position.
  • Offers a fresh alternative to standard Wordle-style puzzles.
Screenshot of Guessir, the word guessing game being played with the player successfully guessing the correct answer of "Jester"