Results, not claims.

Sanitized, results-led summaries from senior engineering engagements. Hard numbers only — no client-confidential detail.

by Willian Pinho — Maintainer of the MCP server that works in Claude Code, Cursor & Gemini.

  • 10M+ users at 99.9% uptime
  • 16y production engineering
  • 3wk KYC pipeline shipped end-to-end
  • 95% verification accuracy
Cross-Platform Payment SDK A major payment platform

The challenge

Deliver a single payment SDK that behaves identically on iOS and Android, holds up under production traffic from 10M+ users, and meets payment-grade security across a fragmented fleet of hardware — without regressions as the platform evolved.

Architecture & contributions

  • Designed and maintained a cross-platform payment SDK in Swift (iOS) and Kotlin (Android) with a shared contract exposing consistent behavior to client integrators.

  • Owned architectural decisions and recorded them as ADRs — giving the client team a durable decision log instead of tribal knowledge.

  • Built the SDK to run reliably across multiple hardware vendors by isolating vendor-specific behavior behind stable interfaces.

  • Mentored 8+ client engineers on the architecture, raising how the team reasoned about payment-critical code.

Results

  • 10M+ users at 99.9% uptime
  • 0 critical security incidents over 12 months
  • 8+ client engineers mentored

SDK remains in production.

Stack

  • Swift
  • Kotlin
  • Payment-grade security
  • ADR-driven architecture
  • Cross-platform SDK
KYC AI/ML Pipeline A fintech KYC/onboarding platform

The challenge

Replace a manual KYC review taking up to 48 hours with an automated, accurate, and cost-efficient pipeline — on a 3-week timeline without introducing production risk at launch.

Architecture & contributions

  • Built an end-to-end AI/ML KYC platform combining TensorFlow (model inference), AWS Textract (document extraction), and OpenCV (image processing) to verify identity documents automatically.

  • Owned the full stack: a TypeScript/Node.js backend on AWS Lambda, with S3 + SQS for auto-scaling document intake and queue-based processing.

  • Secured the API with OAuth/JWT authentication.

  • Wrote customer-facing runbooks so the client team could operate the system independently after handoff.

Results

  • 48h → 5min processing time
  • 95% verification accuracy
  • 70% reduction in operational cost
  • +40% customer satisfaction (CSAT)

Delivered end to end in 3 weeks. Zero critical production issues at launch.

Stack

  • TypeScript
  • Node.js
  • AWS Lambda
  • S3
  • SQS
  • TensorFlow
  • AWS Textract
  • OpenCV
  • OAuth/JWT

Ready to harden your agent stack?

The MCP Gateway Readiness Audit starts with a 15-minute discovery call. No pitch deck — a direct conversation about where your stack is today and whether the audit is the right next step.

Book a 15-min discovery call Other ways to reach me →