Turning fragmented compliance signals into integrated risk analysis.
End-to-end delivery of an analytical framework connecting Internal Audit, Market Security and technical teams inside Brazil's regulated electricity market.
Regulation creates data problems before it creates dashboard problems.
At CCEE, market and compliance decisions sit on top of complex rules, systems and data. Internal Audit, Market Security and technology teams each held different parts of the picture. The analytical challenge was to connect them into a repeatable risk-analysis framework.
Fragmented signals make risk harder to see and harder to act on.
The work required translating risk and compliance requirements into something technically implementable: data definitions, transformations, analytical rules, reporting needs and a structure that could support investigation rather than one-off manual analysis.
Bridge the risk question all the way to the data layer.
Translate risk intent
Worked across Internal Audit, Market Security and technical stakeholders to turn compliance concerns into explicit analytical requirements.
Shape the analytical foundation
Contributed to data modelling and ETL design, using advanced SQL for extraction, transformation, investigation and solution design.
Support implementation
Contributed to SAS analytics implementation and ran proof-of-concepts using Python, Power BI and Tableau.
Make findings usable
Developed executive-level data storytelling and recommendations so analytical findings could influence action.
A measurable reduction in compliance risk.
The 32% figure is the verified internal result associated with the Integrated Risk Analysis framework. The implementation was cross-functional; my role was to lead end-to-end delivery and bridge the business, analytical and technical pieces.
What changes when three warning lists become one ranking.
Not project data. It combines three invented signal lists for invented market participants into one score. The numbers below are computed at build time from invented data made with a fixed random seed. They show how the idea works, not what happened on the project.
| List (each owned by a different team) | Top 5 participants |
|---|---|
| Audit findings | Participant B, Participant D, Participant I, Participant L, Participant S |
| Security alerts | Participant J, Participant L, Participant A, Participant E, Participant M |
| Data-check failures | Participant H, Participant P, Participant T, Participant C, Participant J |
Across the three lists, 13 different participants appear, and only 2 appear on more than one. Each team sees its own slice.
| Combined rank | Participant | Score (0 to 1) | Top-5 lists it is on |
|---|---|---|---|
| 1 | Participant L | 0.89 | 2 |
| 2 | Participant P | 0.73 | 1 |
| 3 | Participant J | 0.70 | 2 |
| 4 | Participant X | 0.66 | 0 |
| 5 | Participant B | 0.66 | 1 |
3 of the combined top 5 (Participant P, Participant X, Participant B) are on at most one single list. A participant can be a little worrying on every list and still never reach the top of any of them.
This does not reproduce the framework, its metrics or the 32% figure above. It only shows why a combined view finds cases that separate lists miss.
How this was computed
24 invented participants. Counts per list are random and independent. Within each list a participant
is ranked, ranks are scaled to 0 to 1 (ties share a rank), and the three scaled ranks are added with weights
audit findings 0.4, security alerts 0.35, data-check failures 0.25. The weights are invented. Code:
src/lib/artifacts/risk-signals.mjs.
Business analysis with technical depth.
I led
- End-to-end framework delivery
- Requirements and stakeholder translation
- Risk / compliance reporting definitions
- Analytical PoCs and executive storytelling
I contributed with specialists
- Data modelling
- ETL design
- SAS analytics implementation
- Technical productionisation
The framework continued into a regulated market-security capability.
I left CCEE in September 2022. Public sources show the framework's later evolution and operating context.
The strongest analysts often work before the dashboard.
The reusable lesson was that decision quality depends on definitions and data structure long before visualisation. My SQL and systems background made it possible to challenge the problem at that level instead of treating analytics as a presentation layer.