Project Ideas

1. English to Terraform

Overview

This project aims to build a system that translates human-readable English descriptions of cloud infrastructure into Terraform code. The generated code can then be used to provision and manage infrastructure resources in an automated manner.

Features

• Translate English sentences into Terraform configurations.
• Support for common cloud providers such as AWS, Azure, and GCP.
• Utilizes GPT-based models like Mistral or Huggingface for natural language processing.

Stack

• Language Model: GPT (Mistral / Huggingface)
• Dataset: Open Source Terraform code available on GitHub

Steps for Development

1. Collect a dataset of Terraform code from open-source repositories.
2. Fine-tune GPT models to generate Terraform code from English input.
3. Evaluate the output for accuracy and correctness.
4. Create a validation system to ensure the generated code is syntactically and semantically correct.
5. Implement deployment features for practical use cases.

Future Enhancements

• Add support for additional cloud providers.
• Integrate a graphical user interface for user input and visualized output.

2. AI Watermarking

• Watermarking refers to the process of embedding a non perceptible to human symbol or set of symbols or patterns to confirm that a given output is generated by a known source
• The goal is to implement an architecture to watermark AI output which can be verified by another algorithm, which should be built as part of the watermarking architecture as well

3. Fastest Way to your University

• Usually, there are multiple ways of transport to university as a student, taking Northeastern and UIC as examples, we have our own university shuttles along with public transportation and local bike sharing like divvy
• We can consume data from apis of university shuttles and apis of public transport (Chicago Transport Authority and Boston for example) we can optimise the path a student can take to reach campus on time
• A unique feature could include co ordinating bus and shuttle switches, google maps does not take into account the arrival time of a connecting bus and does not consume data from university shuttles (like UIC Ride and RedEye for UIC and NEU)

4. Public transport visualisation

• Inspiration from https://github.com/geohacker/bmtc
• Need to check for available APIs/data
• Can start with Chicago/Boston

5. Multi Agent Emergence

• Study emergent behaviour when many agents come together to act as one.
• Simulate an environment where these agents compete for resources
• Further, study emergence between these cumulative agents

6. Pomodoro Timer

• Easy webapp to track progress and time spent on each task, project
• Chrome extension

7. Website for pibbles bakery

• Menu, store locator/ other basic things
• Orders + customisation of cakes
• Chat for conversations on custom orders
• CMS for promos/deals/updates