Project Title: Project Guide Using Agentic Systems 

Hackathon Project Report: Inferentia 

Overview 

During Inferentia, our team developed an innovative web-based platform designed to assist users in selecting relevant problem statements and efficiently distributing tasks among team members. The project leveraged the power of locally hosted Large Language Models (LLMs) and a machine learning approach to enhance collaboration and problem-solving. 

Key Features 

1. Problem Statement Generation: 
2. Users can select a category of interest, and the system suggests relevant problem statements. 
3. Implemented using locally deployed LLMs via Ollama. 
4. Utilised Llama 3.1B, Mistral, and Qwen models to generate diverse and high-quality problem statements. 
5. Intelligent Workload Distribution: 
6. A Random Forest model was trained to analyse the skill sets of team members and match them with the skill requirements of a project. 
7. The model ensured fair and optimised task allocation, improving overall efficiency and collaboration. 

Technical Implementation 

• Backend: Python-based API to handle LLM queries and ML-based task distribution. 
• Frontend: Interactive web interface for category selection, problem display, and team management, using HTML and CSS. 
• Integration: Flask to integrate frontend and backend.  
• Models Used:  
• LLMs: Llama 3.1B, Mistral, and Qwen for problem generation. 
• Machine Learning: Random Forest for team workload optimisation. 
• Deployment: All models were run locally using Ollama to ensure low latency and privacy. 

Challenges & Solutions 

• Optimising LLM responses: Experimented with different models and prompt engineering techniques. 
• Balancing workloads fairly: Fine-tuned the Random Forest model using real-world skill datasets. 
• Efficient local model hosting: Utilised quantised versions of LLMs to optimise computational efficiency. 

Impact & Future Scope 

• This tool can be expanded into a full-fledged AI-powered hackathon assistant, helping teams form ideas and allocate work more effectively. 
• Future enhancements include real-time feedback loops and integration with GitHub or project management tools for continuous tracking.