AI Product Recommendation Chatbot
Budget: $10 – $30 USD
I need a natural-language chatbot that can hold short, human-like conversations and, from those exchanges, suggest the most relevant products in my catalogue. The core of the job is building or fine-tuning an NLP model so that, given a user’s free-text request (“I’m looking for noise-cancelling headphones under $200”), the system returns accurate, explainable recommendations pulled from our existing product database or API.
Here is the flow I have in mind:
1. User greets or describes a need.
2. Bot asks one or two clarifying questions if required.
3. Bot returns up to three product suggestions, each with a brief rationale and a direct link/ID.
4. Conversation ends with an optional follow-up (“Need anything else?”).
You may leverage frameworks such as Rasa, Dialogflow, LangChain, or a custom Python stack with Hugging Face transformers—whatever you feel achieves fast inference and is maintainable. I will supply a CSV/JSON feed containing product titles, categories, attributes, and URLs.
Deliverables
• Trained or fine-tuned model files
• Source code for the dialogue manager and recommendation logic
• A brief README showing setup, environment, and how to add new products
• Short demo video or live link proving end-to-end interaction
Acceptance Criteria
• ≥90 % accuracy on a held-out set of sample queries (I’ll provide these)
• Response latency under 2 seconds on a standard cloud instance
• Clean, well-commented code that I can extend internally
If this matches your expertise in NLP-driven chatbot development, let’s discuss the dataset size, preferred tooling, and timeline so we can get started right away.
Here is the flow I have in mind:
1. User greets or describes a need.
2. Bot asks one or two clarifying questions if required.
3. Bot returns up to three product suggestions, each with a brief rationale and a direct link/ID.
4. Conversation ends with an optional follow-up (“Need anything else?”).
You may leverage frameworks such as Rasa, Dialogflow, LangChain, or a custom Python stack with Hugging Face transformers—whatever you feel achieves fast inference and is maintainable. I will supply a CSV/JSON feed containing product titles, categories, attributes, and URLs.
Deliverables
• Trained or fine-tuned model files
• Source code for the dialogue manager and recommendation logic
• A brief README showing setup, environment, and how to add new products
• Short demo video or live link proving end-to-end interaction
Acceptance Criteria
• ≥90 % accuracy on a held-out set of sample queries (I’ll provide these)
• Response latency under 2 seconds on a standard cloud instance
• Clean, well-commented code that I can extend internally
If this matches your expertise in NLP-driven chatbot development, let’s discuss the dataset size, preferred tooling, and timeline so we can get started right away.