3×
Integration delivery velocity
Transfero · 2023 →
Designed a Microkernel architecture for external-integration adapters and built a Visual Studio template that scaffolds the full library structure. Tripled delivery speed for new adapters and became the team standard.
Trade-off accepted
Accepted more complexity in the core - an extensible kernel costs more to maintain - in exchange for linear adapter delivery. Worth it because integrations were the team's growth vector.
C# / .NET 7MicrokernelClean ArchitectureVisual Studio
Py → .NET
Critical migration from Python to .NET 7
Transfero · 2023 →
Rewrote a critical integration from Python to .NET 7. Performance gains, higher reliability, and alignment with the company's core stack - fewer incidents on critical financial operations.
Trade-off accepted
Accepted the cost of rewriting battle-tested code in exchange for consolidating the stack in a single language - smaller oncall surface, one CI/CD pipeline, unified observability.
Python.NET 7PerformanceReliability
↓ cost
Centralized pricing service
Transfero · 2023 →
Built a centralized service that aggregates external pricing data and serves every internal application. Eliminated redundant calls, standardized the source of truth, and cut company-wide operational cost at scale.
Trade-off accepted
Accepted a single point of failure in exchange for consistency and cost. Covered with aggressive Redis caching, clear SLOs, and graceful degradation when the upstream goes down.
C# / .NETRedisAzureDistributed Cache
SDK
SDK + template that accelerated the payments team
Transfero / BTG Pactual · 2022 →
Built an SDK and a Visual Studio template that scaffold the base project for payment adapters. Other teams now focus only on the integration itself - time per adapter fell from weeks to days, with quality consistent by design.
Trade-off accepted
SDKs demand semantic versioning discipline and an API surface that ages well. I took on that responsibility to remove cognitive load from consumer teams - they no longer need to know the plumbing.
C# / .NETSDK designVisual StudioDX
7 days
Regulatory delivery under pressure
Financial system · delivered in one week
A project with tight regulatory requirements delivered end-to-end in seven days. Prevented significant operational losses and kept the company compliant with the regulator.
Trade-off accepted
Knowingly accepted tech debt - minimal coverage, deferred refactor - in exchange for a non-negotiable deadline. Follow-up plan written on delivery day; debt cleared in the two sprints that followed.
High-pressure deliveryComplianceExecution
CPN
Circle Payments Network integration with DFNS custody
Transfero · 2025 → 2026
Full adapter for CPN (withdrawal, compliance RFI, balance validation) plus the payment flow in the trade engine. In 2026, asynchronous signing through DFNS with approval policies and on-chain transaction V2 (Permit2) behind a feature flag - cross-border stablecoin settlement with institutional custody.
Trade-off accepted
Asynchronous signing adds intermediate states and more error surface than inline signing. Accepted because approval policies require it - and because every step became observable and replayable.
Circle CPNDFNSStablecoinsFeature flags.NET
PIX
Swap Engine: BRL↔crypto on/off-ramp with QR Code payment
Transfero · 2024 → 2025 · my most active repo by PRs
Evolved the company's swap engine: flow resilience and transparency, observability, refund status and flow, unlock endpoints, an idempotency guard per TradeId and the PIX QR Code payment epic. Also worked L3 on ramp incidents - operations fed the reliability backlog.
Trade-off accepted
Strict idempotency per TradeId rejects legitimate retries with a different payload. I chose to reject and alert rather than risk a duplicate settlement - with money, the false negative is cheaper.
On/off-rampPIXIdempotencyObservabilityL3 support
KYT
On-chain compliance with Chainalysis in the crypto engine
Transfero · 2023
Led the Chainalysis blocking epic: alert listening on deposits and payouts, a verification layer with new statuses, release/block endpoints, severity filters, a feature flag and a KYT cron job. Suspicious transactions now stop before money leaves.
Trade-off accepted
Blocking by severity delays legitimate low-risk payouts. Accepted the friction with a manual release queue - predictable operational cost against unpredictable regulatory risk.
ChainalysisAML / KYTComplianceFeature flags