Car-ing is Sharing LLM Chatbot Prototype
OCTOBER 6, 2026 · Career · LLM
After Introduction to LLMs in Python I did a DataCamp project called Car-ing is Sharing, where I played the AI developer for a car sales and rental company that wanted a chatbot prototype. It had to take text and handle a few tasks using pre-trained Hugging Face models.
What I built
All five tasks ran on a small set of car reviews, one Hugging Face pipeline each:
- Sentiment classification with DistilBERT
- Translation of the first two sentences of a review into Spanish with a Helsinki-NLP model
- Question answering with MiniLM ("What did he like about the brand?")
- Summarization of the last review with BART
How it did
I used the evaluation metrics from the course to score it. Sentiment classification got 80% accuracy and an F1 of 0.86 on five reviews, which is a tiny sample. The translation scored a BLEU of 0.78 against a reference. The question answering pulled "ride quality, reliability" out of the review. The summary read fine, though it was cut off mid-sentence because I capped it at 55 tokens.
What I took from it
Each task was only a few lines once I picked the right model, so most of the work was choosing one and checking its output instead of trusting it. It was also the first time I used the metrics from the course on something I built myself.