I work on risk and reporting systems for asset management: VaR and expected shortfall, performance attribution, and the pipelines that produce them on a schedule. MSc in Data Science at NOVA IMS, BSc in Applied Mathematics and Computation at Instituto Superior Técnico.

Project Manager, Risk Team & Horizon Fund · Sep 2024 – Present · live platform

A risk platform for 3 live funds, used daily by 13 fund managers. A 30-module Python engine behind a Streamlit interface, with nightly CI, atomic database swaps, fail-closed data-quality gates and 670+ automated tests.

Risk measures are EWMA filtered historical simulation, GARCH(1,1) with volatility-regime detection, Student-t, and 5-variant Monte Carlo, backtested with Kupiec and Christoffersen. Attribution is Brinson-Fachler with Carino chain-linking. A Black-Litterman rebalancer produces mandate-capped, pre-trade-compliant order lists with transaction costs. Fund commentary is written by an LLM, but every figure in it is injected from code and checked against the source before the report renders.

I also maintain the club's Next.js site, with Vitest on units and Playwright on routing and mobile layout. Both repositories are private.

Quantitative Data Engineer, bachelor's thesis · Feb – Jul 2025 · NASDAQ-listed wealth manager, ≈$75B+ AUM/AUA · sanitized code

Sole engineer on the fund reporting infrastructure. A vectorised NumPy/pandas pipeline computing 21 KPIs across 8 time windows under no-look-ahead constraints took the reporting cycle from hours to under 30 seconds across 200+ funds.

Six Morningstar feeds were consolidated into one point-in-time source of truth with dynamic multi-benchmark mapping, behind Power BI reporting used daily by the investment team and presented to the CIO. Because the output went to regulated review, I also built a Pearson-ρ fund identity detector (ρ ≥ 0.90, p < 10⁻⁹, validated over 20,000 Monte Carlo trials) and regex classification of 175 exposure headers, with 0% unknowns.

Risk Management Intern · Jul – Aug 2024 · CaixaBank Group · alerts · analytics

Daily liquidity monitoring across the fund range. I classified monthly fund flows into 5 movement segments across 19 analytical sheets, producing 20+ client lifecycle KPIs delivered to the CEO of BPI Asset Management.

The manual morning check was replaced by an R and Outlook COM pipeline sending threshold-triggered liquidity alerts before market open, with zero missed reports over the internship. Two further pipelines covered cross-fund 2σ outlier detection and ARIMA(1,1,1) forecasting with 95% intervals.

The two headline numbers, whatever the asset class:

$$\mathrm{VaR}_{\alpha}(L) = \inf \lbrace \ell \in \mathbb{R} : \mathbb{P}(L \gt \ell) \le 1 - \alpha \rbrace$$

$$\mathrm{ES}_{\alpha}(L) = \frac{1}{1 - \alpha} \int_{\alpha}^{1} \mathrm{VaR}_{u}(L) \mathrm{d}u$$

Both depend entirely on the estimated distribution of $L$, which is where filtered historical simulation, a GARCH recursion, a Student-t tail or a Monte Carlo engine come in. Backtesting decides whether that estimate stays in production, which is what the dashed arrow does.

Each repository documents its requirements, its outputs, and the limits of its conclusions. Group coursework credits the full team.

•   MSc, Data Science and Advanced Analytics · NOVA IMS
•   BSc, Applied Mathematics and Computation · Instituto Superior Técnico
•   Quantitative Data Engineer, bachelor's thesis · AlTi Global
•   Risk Management Intern · BPI Asset Management
•   Risk, analytics and web platforms · Técnico Investment Club

Portuguese (native) · English (C1) · Spanish (conversational).