Skip to content
VT
All work
BackendDistributed systems

Async bulk onboarding for a company-wide product merger

Core engineer on a merger with a hard business deadline. Rebuilt account onboarding from synchronous per-row writes into async queued jobs with batched inserts, and shipped the first statistically-gated automation for two new platforms.

When
Mar — Jun 2026
Role
Core engineer
Context
Merging two product lines (Search and Social) into one subscription

Context

The company ran two product lines with separate subscriptions. Leadership set a hard business deadline — June 1, 2026 — to merge them into one, which meant two new platforms had to become first-class in onboarding, automation and the AI assistant.

What I built

Onboarding and account linking

  • OAuth flows for Meta and LinkedIn in the onboarding wizard.
  • Converted account linking from synchronous per-row inserts to async queued jobs with batched bulk inserts.
  • Sanitization for compound Instagram / LinkedIn asset IDs, with defensive rejection of malformed IDs; CSRF handling; a check that at least one account actually landed.

Statistically-gated automation

  • Exclude low-performing placements (Meta) — including Advantage+ → manual placement conversion and guards against removing the last placement.
  • Pause poor-performing LinkedIn Audience Network — pause decisions use rule-of-three confidence bounds and a minimum-click floor agreed in founders’ review, so low-volume entities never trigger false-positive pauses.
  • Social scheduling in the pipeline: prioritize cached accounts that haven’t run in 24 hours.

Assistant support

Exposed the new suggestions to the AI assistant — see AI features in production.

Outcome

The onboarding and Express pillars shipped for the deadline. Express and the assistant became genuinely cross-platform, and the social optimizations now run on 1,000+ accounts and 5,000+ ad sets a month.