We are a startup that has developed a tool for automatic generation of legal texts. Our software is implemented in C#/.NET and is processing input texts with a sequence of finetuned prompts. This is working well and a prototype is already in trial use at several large law firms. We are now looking for someone to work on improving this model. The input is mostly coming from an attorney with machine learning background, who is also providing the texts of the prompts. However, our tasks are complicated and require chaining many prompts in order to get to the desired output.
The output is in same cases simple text, but we will probably more and more need for more sophisticated output such as formatted Word documents or web pages with simple interaction possibilities (e.g. the user should be able to rate and/or edit answers).
You should have previous experience with natural language processing. Knowledge in processing of legal tech and/or creation of Word documents would be great but this can also be learnt later.
We have previously attempted using LangChain and Microsoft Prompt Flow. Our experience at the time was that doing everything in our own code is simpler and more efficient than LangChain. With Microsoft Prompt we had the impression that it is not yet possible to implement loops, which however are required for our processing tasks. If you believe that one of these frameworks would actually be suitable, we would be willing to learn from you. You should be able to understand our existing C# code, but we would be fine with moving to a different language.
We are looking for a long term collaboration that could quickly evolve into fulltime.
This job is already closed and no longer accepting applicants, sorry.
