AI Voice Agent + Outbound Calling Pipeline (CSV → Calls → CSV)
UpworkGBNot specifiedexpertScore: 63
PythonNode.jsAPI IntegrationAutomationBusiness Process AutomationGoogle Sheets AutomationSpreadsheet Automationvapivoice agent
Overview
We need an engineer to build a pilot system that takes a CSV of UK candidate records, places outbound calls via a VoIP provider, runs a short AI voice conversation, and returns an updated CSV with a clear status and short call summary per record.
The goal is accuracy, clean logging, and a reliable weekly run (1,000 calls over ~7 days is fine).
What you will build (high level)
1) Input & Data Handling
* Ingest a CSV (we’ll start with a 50–200 row sample, then scale to 1,000)
* Validate and normalise UK phone numbers (E.164 preferred)
* Deduplicate records
* Apply a suppression / do-not-call list (if provided)
2) Outbound Calling (VoIP)
*Respect UK calling windows
* Throttling / max calls per hour/day
* Handle outcomes: no answer, busy, voicemail policy, invalid number, failed calls
* Optional: retry policy (1–2 additional attempts) if recommended
3) AI Voice Agent (VAPI suggested, but open to other platforms)
* Build the agent using VAPI
*Ask one core question: “Are you currently open to new opportunities / actively looking?”
* Optional minimal follow-up if afctive (keep short)
Produce:
CSV out with a standard disposition/status
* a 1–2 line summary of the call
* optional extracted fields (e.g., location, availability)
4) Outputs & Logging
* Generate an updated CSV for re-import into the client’s system
* Minimum output columns:
* unique_record_id (same as input)
* status/disposition
* short_summary
* last_called_at (timestamp)
* attempt_count
* raw_call_outcome (optional)
* Store a run log (record_id, call id, timestamps, status)
* Optional: store transcripts and/or recording links (if enabled)
Disposition list (pilot-level)
Keep it simple:
* ACTIVE_LOOKING
* NOT_LOOKING
* CALL_BACK
* NO_ANSWER
* WRONG_NUMBER
* DNC
Important constraints
* Accuracy and auditability matter more than speed
* 1,000 calls spread over a week is acceptable
* Deliverables are operational outputs (status + summary), not conversion outcomes
* Clean handling of duplicates, retries (if used), and suppression is essential
Required experience
* Strong VAPI experience (voice agents, call flows, webhooks/events)
* Experience building outbound calling integrations with a VoIP provider (your choice)
* Solid backend skills (Python or Node.js)
* CSV pipelines, phone normalisation, and state/logging best practices (idempotency)
Nice to have:
* Experience with orchestration tools and/or lightweight worker/queue patterns
* Experience deploying to a VPS or cloud with basic monitoring
## What we want from you in your proposal (mandatory)
Please include:
1. Your recommended architecture for this pilot:
For example,
* Python/Node batch runner + webhook receiver
* hybrid
2. How you’ll implement:
* throttling + UK calling windows
* retries (if any)
* idempotency (avoid double-calling)
* logging + output CSV generation
3. A brief description of your VAPI experience (examples/links if possible)
4. Your estimated timeline and cost for:
* MVP on 50–200 sample records
* Full pilot run of 1,000 records
To apply
Start your response with: VAPI VOICE PILOT so we know you read the brief.
Unlock AI Intelligence, score breakdowns, and real-time alerts
Upgrade to Pro — $29.99/moClient
Spent: $924.25Rating: 5.0Verified