Building a calendar-driven CRM and messaging platform
TickMessage is a live Pacifica product at tickmessage.com. It turns calendar activity into personal follow-up, sent from the user’s own WhatsApp and email accounts.
- Status
- Live product, built and run by Pacifica
- Channels
- WhatsApp, Gmail, Outlook and SMS
- Core
- Rails, PostgreSQL, Sidekiq, Go
Situation
Relationship-driven professionals usually know when they should follow up. The follow-up still depends on memory and a task list, so it often happens late or not at all.
The product had to send that routine follow-up automatically without it reading like a bulk campaign, and without moving the user to an unfamiliar sender identity.
Responsibility
Pacifica owns TickMessage. Product design, engineering, infrastructure and day-to-day operation are all Pacifica’s work, with Tarun Mookhey as principal engineer.
Work delivered
Product
- Calendar rules with offsets and filters that turn events into triggers
- Contact resolution that matches the right person to each event
- Reusable templates with merge tags
- Delivery through WhatsApp, Gmail or Outlook, and SMS
Operations
- Scheduling and recurrence for repeat follow-up
- Attachments held in S3 behind presigned URLs
- Health checks, back-off, recovery and alerting
- Unit, system and browser-level tests
Outcomes
TickMessage is in production. Users keep their existing messaging identity, so replies return to the real conversation rather than a marketing inbox, while scheduling, rules, delivery state and recovery run asynchronously underneath.
Technical details
Architecture
A Rails and PostgreSQL core holds contacts, rules, schedules and delivery state. The WhatsApp connection runs as a separate Go service, so a stateful messaging session cannot destabilise the web application.
Delivery and search
- OAuth2 across Google Calendar, Gmail and Contacts; messages send as the user
- Sidekiq with cron scheduling and a recurrence engine
- Redis streams with consumer groups, with normal and bulk priority queues
- Trigram, full-text and vector search in PostgreSQL
Have a workflow that depends on several systems behaving as one?
Tell us where identity, scheduling, third-party APIs or background processing are creating complexity.