01FINANCIAL LEADERSHIP
25+ years
CFO-level financial judgment
FP&A · valuation · M&A · investment decisions
AI EVALUATION VALUE
Test whether AI conclusions make financial sense.
FINANCE AI MODEL EVALUATION · EXPERT REVIEWBRAZIL · INTERNATIONAL REMOTE
JOÃO MARCELO ALVES / SENIOR FINANCE EXECUTIVE
A convincing AI answer isn't enough.
It has to withstand scrutiny.
I bring 25+ years of CFO-level financial judgment to evaluating AI-generated analysis—reconciling the numbers, challenging unsupported reasoning and making evidence-based scoring decisions.
02EVIDENCE & DILIGENCE
M&A due diligence
Source-driven analysis
Financial reconciliations, critical assumptions and reporting integrity
AI EVALUATION VALUE
Detect unsupported claims and source inconsistencies.
03AI FINANCE DESIGN
FDI + Sentinel
Built evaluation-ready frameworks
LLM-assisted analysis and a 30-test accounting control framework
AI EVALUATION VALUE
Define ground truth, rubrics and critical-failure checks.
01 / THE EVIDENCE
Judge my work.
Follow the evidence.
Two finance AI evaluation cases built around observed development outputs. Each makes the task, reference evidence, scoring logic and expert adjudication inspectable.
CASE 01FINANCIAL REASONING
T4R Financial Decision Intelligence
When cash-flow reasoning sounds right, what does the evidence prove?
Review cash-flow reconciliation and runway claims against recorded figures, classifications and assumptions. See where plausible results must be qualified or held.
Cash-flow bridge: HOLDRunway scenario: CONDITIONAL USE
CASE 02ACCOUNTING CONTROLS
T4R Sentinel Control
When AI reports control coverage, how much assurance is justified?
Review observed test coverage, source lineage and independent verification status. See why an accurate count is not the same as full accounting assurance.
Reported scope: LIMITED ACCEPTANCEEvidence verification: HOLD
Evidence standard: the cases analyze recorded historical development outputs. Their evaluator dispositions and scoped scores are simulated expert adjudications—not controlled benchmark results for named LLMs.
02 / HOW I EVALUATE
One framework.
Different financial AI workflows.
I use an evidence-first, five-step workflow across financial AI models. The task, source data and scoring criteria adapt to each use case; the standard of proof stays consistent.
Define the test
Identify the model’s task, intended decision and material financial failure modes.
DELIVERABLE Evaluation task & risk scope
Establish ground truth
Reconcile source records, reproduce key calculations and define evidence limits.
DELIVERABLE Verified reference & sources
Design the rubric
Specify observable checks, weights, acceptance thresholds and critical-failure gates.
DELIVERABLE Scoring rubric & gates
Test the output
Run documented test cases, preserve raw outputs and compare claims with the reference.
DELIVERABLE Execution record & findings
Score & adjudicate
Apply evidence-backed scores; identify critical errors, decision risk and release conditions.
DELIVERABLE Scored judgment & next actions
ONE STANDARD. ADAPTED TO THE TASK.
Keep the evaluation sequence; tailor the inputs, rubric and failure gates.
FINANCIAL REASONING / T4R FDI
Cash-flow evidence · Assumptions · Decision quality
ACCOUNTING CONTROL / SENTINEL
Reconciliation · Test coverage · Source integrity
03 / THE DOMAIN EXPERT
Financial experience is my evaluation advantage.
I have been accountable for financial reporting, investment analysis and executive decisions. That experience tells me which AI errors matter, which conclusions overreach the evidence and when a result is not decision-ready.
SELECTED FINANCIAL TRACK RECORD
01
M&A and due diligence
Supported reporting readiness and financial-evidence review for a BRL 48.7M transaction.
02
Capital and investment decisions
Supported financial strategy for BRL 14M in equity and innovation financing.
03
Reporting discipline
Reduced a financial reporting cycle from 50 to 20 days through process and system integration.
My evaluation standard: Can I reproduce the number, defend the reasoning and trust the decision?
04 / LET'S TALK
Hiring someone to evaluate financial AI?
Bring me the hard cases.
Finance AI model evaluation · Expert review · English (C1) / Portuguese · International remote