Skip to work
Back to work
02 — AI Content Platform · Contractor Website Services

CWS — Voice AI Platform

Replaced static intake forms with phone conversations for a platform used by 200+ people.

Role

Associate Software Engineer — frontend & backend

Technologies
ReactNestJSRetell AIRedisBullMQ

An AI content platform for contractors, used by 200+ people. I integrated the five Retell AI voice agents that replaced intake forms, plus the queue that schedules their calls.

voice-agent / call-flow
  1. 1Greeting · AI disclosed
  2. 2Intent: reschedule
  3. 3Check availability · API
  4. 4Confirm & close
Concept illustration

Note — Client project — no confidential code or data is shown. Visuals are concept illustrations.

00 / How a call moves through the system
  1. 01
    Scheduled

    A callback job is queued in BullMQ on Redis.

  2. 02
    Voice agent

    One of 5 Retell AI agents runs the conversation.

  3. 03
    Tracked

    Call status is recorded end to end.

  4. 04
    Reviewed

    Mission Control shows agent activity; AI actions are reviewed before going live.

  5. 05
    Calendar

    Team calendar shows calls, outcomes and durations.

When things fail
  • A failed call is not lost — the queue retries it automatically.
  • Status tracking makes it visible which calls succeeded or still need attention.
  • Humans check AI actions in Mission Control before they reach production.
01 / Motivation

Static intake forms were slow and impersonal. Contractors needed a faster way to discuss content, verify details and approve work.

02 / Solution

Retell AI voice agents handle content discussions, verification, approvals, service mapping and geographic data collection over the phone — backed by a queue-driven callback system and a live review dashboard.

03 / My contributions
  • Integrated 5 production Voice AI agents with Retell AI
  • Built automated callback scheduling with Redis and BullMQ, with retries and end-to-end call status tracking
  • Built a calendar dashboard of team members' daily calls and meetings with client details, outcomes and durations
  • Built Mission Control — a live dashboard to monitor AI agent activity and review AI actions before they go live
04 / Highlights
5 production voice agents200+ usersCallback schedulingAutomatic retriesCall status trackingMission Control dashboardTeam calendarAI disclosure & consent
05 / Decisions & challenges
  • Human review before AI actions go live (Mission Control) keeps automation accountable.
  • Queue-based scheduling with BullMQ so failed calls retry automatically instead of being lost.
06 / Learning

Useful AI needs reliable plumbing around it: queues, retries, and a place for humans to check its work.

Next project
PVC Directory
[09] / Contact

Write a quick note

Prefer to start here? Your note opens in your own email app, ready to send.

Nothing is sent until you press send there.