Icarus the AI Resume Chatbot

After much deliberation on stack, architecture and every other tradeoff, this project lives! Locally usable and can be deployed to production.

Icarus is an AI-powered chatbot that represents me in basic tech interview recruiting Q&A scenarios. Icarus can answer questions about my technical skills and past experience as a software engineer, powered by Retrieval Augmented Generation (RAG). Check out the video to see Icarus answering basic technical recruiting type questions.

Icarus is written primarily in Python, using the LangChain framework for AI powered applications to integrate the application’s architectural components. The application features a server running AI inference, scripts for processing the bot’s internal database to enable inference and a vector data store (Postgres) via which the bot performs document retrieval. The chatbot pulls information from the resume database to retrieve the proper data to answer questions.

Icarus uses Ollama to run AI inference. Ollama is an open source AI model runner that can be used for local development and in the cloud for production instances.

Hugging Face provides access to both AI models used in the application. For the initial load of my resume data into the vector store database, I used Sentence Transformers’ all-MiniLM-L6-v2 model on Hugging Face. This model is part of the generation and processing of embeddings of the original data.

The model used for performing inference for the main chat is Google’s Gemma3. This model goes into action each time a question is asked of the chatbot in order to query the database, retrieve relevant documents and generate an acceptable answer.