FinTech onboarding platform
40% faster KYC
Average time from sign-up to verified account dropped from 11.2 minutes to 6.7 minutes within eight weeks of launch.
A regulated FinTech startup needed to replace a brittle onboarding monolith without pausing growth. We rebuilt the flow as event-driven services — faster for users, auditable for compliance.
- Client
- Series A FinTech · KYC & identity
- Duration
- 16 weeks
- Stack
- Laravel · AWS · PostgreSQL · Redis · SQS
When peak signup days started producing support queues instead of activated accounts, the product team knew the onboarding experience had become a growth ceiling. Users were dropping during document upload, operations could not see where cases stalled, and every compliance tweak required a full release train.
Outcomes at a glance
Median time-to-verified dropped sharply after removing duplicate fields and async document processing.
Status notifications and clearer error copy reduced "where is my application?" contacts.
Feature flags let policy changes ship in days instead of multi-week regression cycles.
Every reviewer action and system decision logged with policy version metadata.
The challenge
The existing flow was a single Laravel application with tightly coupled steps: phone verification, PAN capture, address proof, liveness check, and manual review. Product marketing had added steps over two years without revisiting the overall journey, so users re-entered data and hit dead-ends when third-party verification APIs slowed down.
Compliance needed immutable audit logs — who viewed which document, when a decision was made, and which policy version applied. The legacy system logged inconsistently, which made quarterly audits painful.
Engineering wanted feature flags and incremental rollout. Shipping a new OCR provider or an extra verification step meant weeks of regression testing because everything deployed together.
Our approach
We ran a two-week discovery sprint with product, compliance, and support leads. Together we mapped the happy path, failure paths, and regulatory must-haves into a service blueprint. That became the backlog: modular steps, each owned by a small service with a clear contract.
Identity capture, document validation, risk scoring, and ops review became separate Laravel services communicating through SQS events. The user-facing app orchestrated steps but did not own business rules — so compliance could update a step without redeploying the entire funnel.
We added an operations console where reviewers see queue depth, SLA timers, and one-click approve/reject with mandatory reason codes. Every action streams to an append-only audit store on PostgreSQL with nightly exports to the client's compliance bucket on S3.
How we delivered it
Weeks 1–2 · Discovery
Journey mapping, integration inventory, compliance checklist, and parallel-run architecture decision.
Weeks 3–8 · Core rebuild
Shipped identity + document steps behind feature flags while legacy flow still handled production traffic.
Weeks 9–12 · Ops tooling
Reviewer dashboard, SLA alerts, audit exports, and runbooks for support.
Weeks 13–16 · Cutover
Gradual traffic shift by cohort, monitoring drop-offs per step, hotfix window for edge cases.
Technical decisions that mattered
We chose PostgreSQL for transactional state and Redis for ephemeral session data and rate limiting on verification APIs. SQS gave us retries and back-pressure when upstream identity providers throttled during peak hours.
The frontend stayed a thin Next.js shell: step components fetched state from a BFF layer that hid service fragmentation from the client. That kept mobile web performance acceptable on mid-range Android devices common in Tier-2 cities.
- Feature flags per onboarding step and per OCR provider
- Idempotent webhooks from verification vendors
- Staged rollouts with automatic rollback on error-rate spikes
- Synthetic monitoring simulating full KYC every fifteen minutes
What we would do again
Parallel-running legacy and new flows de-risked cutover. Support and ops trained on the new console two weeks before traffic moved — that prevented the usual "tooling surprise" on launch day.
"We finally ship policy changes without holding our breath. Ops sees the same truth engineering sees — that alone was worth the project."
Planning something similar?
Tell us about your constraints, timeline, and what success looks like. We will share an honest assessment — including whether we are the right fit.