01
Money movement has too many exceptions.
Reconciliation backlogs, ambiguous transaction state, partner failures, and retries consume engineering and operations time.
US payments and cross-border fintechs
Find the failure modes in your money movement or production AI workflow. Start with a paid diagnostic, then agree the implementation your evidence supports.
The inflection point
When money movement or an AI workflow fails, the consequence crosses engineering, operations, and customer support. Trace the selected workflow together, establish the baseline, and decide which controls deserve implementation.
01
Reconciliation backlogs, ambiguous transaction state, partner failures, and retries consume engineering and operations time.
02
Retrieval failures, inconsistent answers, unsafe tool access, latency, and cost lack a shared evaluation baseline.
03
A new corridor, partner, or AI workflow is raising the stakes. Your team needs a bounded diagnosis and a clear implementation decision.
Relevant operating proof
Operating leadership across cross-border payments, global-scale distributed infrastructure, US–LATAM team building, LLM systems, and public open-source work on AI-agent safety.
Read the fintech scale case →$250M+
monthly payments volume supported
Mauricio / Félix context
99.999%
global infrastructure reliability
Francisco / Meta context
40+
person engineering organization led
Mauricio / Félix context
Millions/sec
global CDN request scale
Francisco / Microsoft context
Two reliability offers
Primary offer / money movement
A paid diagnostic for transaction state, retries, reconciliation, settlement exceptions, and partner failures. Establish the baseline and the scope for implementation.
→02Primary offer / production AI
A paid diagnostic for LLM, RAG, or agent workflows. Inspect evaluations, retrieval, failures, latency, cost, and security boundaries before agreeing implementation.
→How the work starts
Use a fit call to agree the workflow, the operating problem, and whether a paid diagnostic is useful.
01
Agree a bounded workflow, inspect representative evidence, establish a baseline, and prioritize the decisions and controls.
02
Agree a separate scope, owners, acceptance measures, and rollout boundaries for the changes the diagnostic supports.
03
Separately agree recurring review, reliability measures, and operating responsibilities when ongoing support is useful.
Pricing, timing, and deliverables are agreed in a proposal. If the question spans architecture, delivery, team structure, and AI readiness, explore the broader engineering assessment →
Company leadership
Fintech Field Office brings senior operating experience across payments, engineering organizations, distributed systems, reliability, performance, and production AI. Clients receive one accountable scope, an integrated recommendation, and clear ownership of the decisions that follow.
Leadership 01
Engineering Principal — Fintech, Payments & Engineering Scale
Fintech engineering leader with 13+ years across payments, credit, distributed systems, data platforms, hiring, and US–LATAM engineering organizations. He currently serves as Head of Engineering at Winston Artory Group after operating roles at Félix Pago, Nubank, Lyft, Oracle, and Nextiva.
Leadership 02
Engineering Principal — Internet-Scale Systems & AI Infrastructure
Francisco has spent 18+ years engineering systems at a scale few teams ever encounter. Now a Senior Software Engineering Manager at Amazon, he previously built and led infrastructure at Meta and Microsoft that served billions of users and handled millions of requests per second. At Félix Pago, he applied that depth to production LLM evaluation, RAG, observability, and AI performance.
Employment history identifies relevant individual experience and does not imply endorsement of Fintech Field Office by any current or former employer.
Start with evidence
Interactive diagnostic / 8 dimensions
Identify the weakest operating dimension and get a prioritized next-step recommendation. No email wall.
Run the scorecard →Complete example / fictional company
Inspect the evidence, eight-dimension score, six prioritized findings, and owned 90-day plan.
Open the report →Public proof / AgentLint
A concrete look at 77 public rules for quality, security, and infrastructure safety—without an invented velocity claim.
Examine the proof →Evergreen decision library
Original guidance on fractional leadership, technical diligence, payment-system scale, and production AI infrastructure—written from the operating evidence behind the practice.
Choose diagnosis or embedded leadership from the decision at stake.
Read →Prepare architecture, controls, evidence, and ownership.
Read →Review invariants, reconciliation, reliability, and operations.
Read →Evaluate quality, retrieval, observability, cost, and security.
Read →Inspect public guardrails for permissions, quality, security, and completion.
Read →What the engagement model and investment should actually buy.
Read →Selected field work
A small portfolio of public tools and product experiments built to test engineering, distribution, and workflow ideas in the real world.
Open source / AI engineering
Public guardrails for safer AI-assisted software delivery.
68k downloads · 200+ GitHub stars
Pepy / GitHub · checked November 3, 2026
View on GitHub ↗Open source / AI agent reliability
Define what “done” means for your coding agent. Verify its work against tests, quality checks, and review criteria—with bounded repair loops when checks fail.
Works with Claude Code, Codex, and Gemini CLI.
View on GitHub ↗Browser game / AI experiment
A playful take on the AI industry: upgrade hardware, generate tokens, and build your little empire. Made for the moments when your coding agent is still working.
A side project by Mauricio.
Play the game ↗