Need a NN model with architecture capable of learning a NLP -> Regression problem

I am looking for a solution, a model that should be able to learn and solve a natural language processing (NLP) to regression problem. The problem is to find the correlational score (matching score) of two contexts. First context will be a text about a person, like person's CV, with information about experience, educations, etc.. Second context will be the job description text, defining the searching criteria. The model gets those 2 contexts as input, and then output a number in range -1 to 1 for matching score. Please do not think of this as a similarity problem, because it does not work. I also tried to build my own models using different techniques, for example a siamese model will give me a loss less than 0.01, but in fact the model will perform wrong after training. For example, the job description text mention that we are looking for a person located in Europe, it should not look only for similar words in two texts, but also reason about the context and learn keyword meaning like job location, where Europe means any country from Europe, so in this case similarity is not working properly, because the location will be different within Europe. This is an dataset example generated with OpenAi's Chat GPT: [login to view URL] The model should be able to work with small datasets, which has less than 1000 samples. The model should be able to be trained and used on google colab which have 16Gb GPU T4. Please bid only if you can provide a solution or if you already made something similar, in order to avoid time wasting. Because somebody may be confident at the beginning then he could fail at the end. So to avoid time wasting, just bid if you can show me something similar and demonstrate that the model is able to solve such problem. Also, please bid within budget range to be considered. You can use open sources from huggingface or whatever works.