ReverieHacks ML prompt engineering

ReverieHacks ML prompt engineering is a track where teams design and orchestrate AI workflows to achieve the best possible results for specific purposes. Participants will focus on creating efficient and effective workflows consisting of several agents like LLM models and human agents with structured queries to ensure the most optimized output based on the given input.

Participating

Building your project

The first step to participating comes through identifying the context of your problem. Many of you might already have a good idea of what the contexts of your problem are, however for those who are still unsure, we have included a problem statement template that can help you get started.

After you have identified the problem, you need to identify your input and output. After that is designing the solution to reach the output from the input in the most efficient way. This can be done on different LLM orchestration apps like Laminar, Vellum and FlowiseAi that are used to design workflows with node actions that are enacted by different systems humans or ML agents. You can mix and match multiple nodes nodes in a chain with specific queries to achieve your purpose.

After designing your algorithmic solution is a repeated process of testing out the algorithm and optimizing it to create a generic workflow that can be reused for achieving optimum results from similar inputs.


Submitting your project

The uploading process can be carried out easily on Devpost under the respective track. Simply upload the following files for a successful submission

ML workflow :

Upload a png of the flowchart version of your ML workflow that depicts where human inputs is necessary, what queries are used with LLMs, which LLM model is used and what each action does.

Samples :

Prepare a video/document that shows your workflow working for sample test cases.

Documentation :

Include a detailed documentation that explains the each node, including its config, model used, cost analysis and any other necessary data.

Guidelines

Guidelines for building your project

These are guidelines that we have added that you must abide by to ensure a fair competition for all. To reach out to us, you can join the discord or reach out to us on our email

  • Human input can be added by combining a node with a human agent. The lesser the human input is used, the better the workflow becomes.

  • If you add LLM agents to a node, then you must also include the cost analysis for that node.

  • All submissions must be the original work of the team. Plagiarism or any form of cheating will result in disqualification.

  • Teams can use open-source libraries and tools but must provide proper attribution.

  • Mentors will be available to provide guidance and support throughout. Teams are encouraged to seek help from mentors

  • All projects must be submitted by the designated deadline. Late submissions will not be considered unless there are extenuating circumstances

  • You can find us on our discord regarding any complaints, feedbacks and sugestions!