Skip to content

Latest commit

 

History

History
28 lines (22 loc) · 2.23 KB

TASK.MD

File metadata and controls

28 lines (22 loc) · 2.23 KB
Raw

Key skills in the GenAI area for now are:

  • Prompt engineering
  • Understanding the importance of the context and techniques working with it
  • The usage of accelerators and integration of various pieces of code (both EPAM and outside)

Task

Demonstrate your understanding of those on a basic level and to deliver a Recommendation system based on GenAI.

  1. Choose any public API or any API you have access to. For inspiration, you can use https://github.com/public-apis/public-apis?tab=readme-ov-file, https://github.com/ozlerhakan/mongodb-json-files/blob/master/datasets/books.json or https://github.com/emilydumas/mcs260fall2021/blob/main/samplecode/data/episodes.json, and grab a couple of dozen items of it.
  2. Build a prompt that will guide a user through the process of choosing an item (like choosing a movie, a city, or whatever) and help him with possible doubts.
  3. [Advanced & Optional]* Integrate it with the EPAM accelerator (https://epam-rail.com/) as a DIAL chat application with a system prompt designed on a previous step. Should be run locally. Please note that this accelerator is well-documented and contains examples - so it is not expected to be a guided exercise.

*The advanced & Optional part is for you if you want to target a more experienced team to challenge yourself and go deeper during the course.

What Good prompt should include:

  • Recommendation logic should be based MORE on the knowledge within the prompt, and LESS on GPT General knowledge.
  • The bot instructed with this prompt should talk professionally, and make an impression of a "salesperson in a shop" and guide a user through the process of choice.
  • For key types of messages, a bot should give structured and predictable responses (like comparing two movies or showing one item). Think of it as it is prepared to be rendered in special widgets.
  • [Advanced] For certain use cases should be even pre-defined "text scripts" within a prompt.

The https://chat.lab.epam.com can be used as a chat engine, but since you're not working with private data here feel free to use public ChatGPT or any other LLM of choice.

Deliverables expected:

  • Prompt
  • [Advanced & Optional] short video of a conversation with your prompt via your DIAL App launched locally and the code of it.