Challenge Overview
Prizes
- 1st Place: $4,000
- 2nd Place: $3,000
- 3rd - 8th Places: $500
Challenge Background
Topcoder has delivered a LOT of solutions since 2001 - software developed by the Topcoder Community flies on the International Space Station, aids cancer treatment, has given robots binocular vision, helped forecast spread of disease, helped ensure data privacy, provided Web3 solutions, helped satellites communicate, discovered and tracked asteroids, helped insurance companies, dentists, doctors, shipping companies, manufacturers - the list is long. So too are the technologies used - mobile, full-stack, data science, cloud, C++ to COBOL to Haskell to Rust to no-code, and so on.
Describing Topcoder capabilities succinctly is hard!
Sales teams and website visitors who want to find the “right” Topcoder case studies for their interests struggle with traditional search methods. Keyword-based searches often miss important contextual information, leading to a suboptimal user experience. We think GPT can help. A conversational chatbot powered by large language models (LLMs) can address this challenge by better inferring the intent behind user inquiries and providing personalized, nuanced responses.
In this challenge, contestants will develop a conversational chatbot using GPT technology to effectively search, analyze, and present Topcoder case studies to sales and website visitors. The chatbot should be able to infer the nuances of user inquiries, identify relevant case studies even when keywords don't match exactly, and provide insightful responses beyond mere keyword matching.
But there’s a catch or two - like many firms today, Topcoder doesn’t just yet want to expose the dataset to OpenAI, or other online LLMs and AI suites. Contestants must use disconnected / down-loadable LLMs (like LLAMA2, but there are others) and libraries, or custom code. Also, the solution will have to sort out a complex and rather cluttered dataset to find the relevant documents for training, and function properly WITHOUT requiring Topcoder to re-organize datasets, create new ones, or in any way develop new data practices.
Challenge Objective
The goal of this project is to develop an LLM-powered conversational chatbot that can:
- Effectively search and retrieve relevant Topcoder case studies based on user inquiries.
- Infer the nuances of user inquiries, even when keywords don't match exactly.
- Provide insightful responses that go beyond mere keyword matching, including personalized recommendations and contextual insights.
- Accurately explain Topcoder capabilities conversationally, combining the case studies provided to this challenge with the Topcoder body of knowledge latent in LLMs with billions of parameters.
- Accomplish these objectives without requiring the data to be cleaned up, moved, labeled or categorized!