looking for AI expert to fix chat bot urgent
Budget: $10 – $30 USD
My production chatbot has started missing the point of conversations—it stalls or returns a fallback message whenever the user asks anything even mildly off-script. The root problem sits in its natural-language-understanding layer: intents are misclassified, entities aren’t picked up, and the confidence scores drop to zero far too often.
The codebase is written in Python and currently relies on a transformer-based model served through a REST API, with a small Rasa NLU pipeline supervising intent detection. I’ve already confirmed the webhook and integration endpoints are healthy; this is squarely an NLU issue.
What I need from you:
• Inspect the NLU pipeline, training data, and model configuration.
• Identify why the model fails to map typical user utterances to the correct intents.
• Retrain or fine-tune the model (TensorFlow / PyTorch is fine—whichever you think best) and adjust confidence thresholds so responses trigger correctly.
• Hand back an updated codebase, fresh training data (if added), and a short test script that proves at least five varied user queries are now handled accurately.
Please keep the turnaround tight—the bot is live on our site and every hour of misfires costs us conversions.
The codebase is written in Python and currently relies on a transformer-based model served through a REST API, with a small Rasa NLU pipeline supervising intent detection. I’ve already confirmed the webhook and integration endpoints are healthy; this is squarely an NLU issue.
What I need from you:
• Inspect the NLU pipeline, training data, and model configuration.
• Identify why the model fails to map typical user utterances to the correct intents.
• Retrain or fine-tune the model (TensorFlow / PyTorch is fine—whichever you think best) and adjust confidence thresholds so responses trigger correctly.
• Hand back an updated codebase, fresh training data (if added), and a short test script that proves at least five varied user queries are now handled accurately.
Please keep the turnaround tight—the bot is live on our site and every hour of misfires costs us conversions.