Customer Support Answering AI Training

The goal is to build a domain-specific language model that can read an incoming customer question and instantly generate an accurate, helpful reply by drawing only on our own support tickets and past chat conversations.

I will provide a sizable, well-structured export of those tickets and chats (text format, English). You will handle every stage of the training workflow: cleaning and anonymising sensitive data, converting the conversations into a Q&A-friendly format, fine-tuning or building the model, and validating its answers against a reserved test set. OpenAI, Hugging Face Transformers, or a similarly proven framework is acceptable as long as the final model can be called through an API or lightweight service that I can later host on my own infrastructure.

Deliverables
• Pre-processed, fully documented training dataset
• Trained language model (weights and config) capable of answering customer support questions with high precision
• Inference script or minimal API endpoint for real-time use
• Evaluation report showing metrics on the held-out test set and examples of correct and flagged answers
• Clear hand-off notes so I can continue iterative training in-house

Acceptance will be based on reproducibility of the training pipeline and the model achieving reliable, context-aware answers that align with the content of the provided tickets and chats.