AI Sales Call Analyzer
Budget: ₹12,500 – ₹37,500 INR
I want to automate the way our team learns from recorded sales calls. The system you build should let users upload an audio file, run speech-to-text, and then surface the parts that actually matter—customer interest, sentiment, and the concrete next steps we agreed on. I’m fine with a moderate transcription accuracy at this stage; perfect word-for-word output isn’t critical as long as the summary logic still works.
Here is the flow I’m picturing:
• Node.js backend receives an uploaded call (MP3, WAV, or similar)
• Whisper, Gemini, or another AI API produces the transcript
• A summarisation layer distills the conversation and tags segments that reflect buyer enthusiasm or hesitation, then pulls out agreed action items
• Results are written to MongoDB and displayed in a simple dashboard rather than a plain text file or raw JSON—think quick visual scan for reps, with an export option if we ever need it
For acceptance, I should be able to:
1. Upload a sample call and see a timestamped transcript inside the dashboard.
2. View an autogenerated summary block that calls out a) customer interest & sentiment and b) next steps and action items.
3. Download or query the same data as structured JSON if I choose.
If you already have experience chaining Node.js with Whisper or a similar model and presenting results through a browser-based dashboard, this should feel familiar. Let me know how you would wire the services together, any off-the-shelf libraries you’d rely on, and a rough timeline for a working MVP.
Here is the flow I’m picturing:
• Node.js backend receives an uploaded call (MP3, WAV, or similar)
• Whisper, Gemini, or another AI API produces the transcript
• A summarisation layer distills the conversation and tags segments that reflect buyer enthusiasm or hesitation, then pulls out agreed action items
• Results are written to MongoDB and displayed in a simple dashboard rather than a plain text file or raw JSON—think quick visual scan for reps, with an export option if we ever need it
For acceptance, I should be able to:
1. Upload a sample call and see a timestamped transcript inside the dashboard.
2. View an autogenerated summary block that calls out a) customer interest & sentiment and b) next steps and action items.
3. Download or query the same data as structured JSON if I choose.
If you already have experience chaining Node.js with Whisper or a similar model and presenting results through a browser-based dashboard, this should feel familiar. Let me know how you would wire the services together, any off-the-shelf libraries you’d rely on, and a rough timeline for a working MVP.