Client overview
The client is a trusted data solutions partner headquartered in the Netherlands. The company specializes in supporting all stages of AI development, from training to evaluation, by delivering high-quality data through services such as data annotation, generation, and collection.
Business Challenges
-
The client faced difficulties in seeking a partner who could:
- Provide a large number of domain experts.
- Handle a high volume of tasks within a short period.
- Meet the requirements for the naturalness of the code.
The client was looking for a vendor offering data labeling services for LLMs to empower businesses with high-quality SFT datasets, enabling the development of safe, responsible, and trustworthy AI products.
Project Detail
Language: Python, C/C++, Java, JS, Scala, .Net, Bash, R
Solutions
Upon receiving the project, LTS GDS took the following steps:
-Analyzed the customer’s problem and clarified their requirements.
-Refined the existing response data (including questions and answers) to ensure natural input for theLLM.
-Generated answers to new questions using technological knowledge and logical reasoning
This process involves:
- Based on the provided Text2code, rewrite questions in natural language, write logically correct responses with explanations, and create test cases as standards.
- Divide the code data from Step 1 into two parts:
-
- Create questions
- Create responses Refine both parts to fit the context and add detailed explanations to the responses.
- Add comments and docstrings to the code generated in Step 1.
- Create a question in natural language, then ask the AI to generate a unit test, answer, and explanation.
