Theme
Finance was chosen because the target role asks for data science in a financial environment, not only generic modelling.
PySpark · Spark SQL · Financial analytics
Financial risk signals from company statement data.
A compact portfolio project that simulates how a data scientist can turn financial data into risk prioritisation, business insights and a simple predictive model.
Executive summary
This is the type of output a business user would need first: what requires attention, where the exposure is concentrated and why the alert exists.
Current findings
- 6 entities are flagged as high risk in the latest quarter, representing €1,085.7m of exposure.
- Payments is the main concentration to review, with €487.2m of high-risk exposure.
- 45 entities show recent deterioration in margin, leverage or debt-service capacity.
Why this portfolio case
The topic was selected because it connects technical PySpark work with a realistic financial-services decision: prioritising entities for review based on risk and exposure. Each step mirrors a common workflow in analytics teams.
Historical data
Real teams monitor evolution over time: margin compression, rising leverage and worsening coverage matter more than one isolated number.
Exposure
Risk only becomes business-relevant when combined with financial impact: where is the money at risk?
Explainable score
The score translates several financial signals into a simple 0-100 view that a non-technical stakeholder can prioritise.
PySpark + SQL
The workflow mirrors production analytics: scalable transformations first, business cuts and summaries with SQL afterwards.
ML baseline
The predictive layer adds prioritisation while the rule-based drivers keep the analysis understandable and auditable.
Sector concentration
Spark SQL aggregates the latest quarter by sector. The coloured bar is the average stress score on a 0-100 scale, so the riskiest sectors are visible at a glance.
PaymentsStress 24.8/100
- Entities
- 20
- Total exposure
- €4,617.7m
- High-risk exposure
- €487.2m
- Coverage
- 3.40x
LendingStress 19.2/100
- Entities
- 20
- Total exposure
- €3,445.4m
- High-risk exposure
- €487.1m
- Coverage
- 5.26x
BankingStress 18.8/100
- Entities
- 20
- Total exposure
- €4,765.8m
- High-risk exposure
- €86.9m
- Coverage
- 4.78x
Asset ManagementStress 16.3/100
- Entities
- 20
- Total exposure
- €4,657.3m
- High-risk exposure
- €24.5m
- Coverage
- 5.65x
InsuranceStress 10.6/100
- Entities
- 20
- Total exposure
- €4,722.2m
- High-risk exposure
- €0.00m
- Coverage
- 7.26x
Entity watchlist
The watchlist combines exposure, an explainable financial stress score, model probability and plain-language risk drivers.
| Entity | Exposure | Stress score | ML probability | Tier | Main drivers |
|---|---|---|---|---|---|
| Meridian Bank 10Banking · 2025Q4 · rating B | €86.9m | 72.7/100 | 65.2% | High | High leverage, Weak interest coverage, Low liquidity, Margin deterioration |
| Horizon AM 08Asset Management · 2025Q4 · rating B | €24.5m | 70.7/100 | 46.0% | High | High leverage, Low liquidity, Negative revenue growth, Margin deterioration |
| EuroMerchant 12Payments · 2025Q4 · rating B | €98.5m | 64.9/100 | 67.0% | High | High leverage, Low liquidity, Negative revenue growth, Margin deterioration |
| CapitalFlex 09Lending · 2025Q4 · rating B | €181.6m | 61.8/100 | 41.7% | High | High leverage, Low liquidity, Negative revenue growth |
| Atlas Payments 16Payments · 2025Q4 · rating BBB | €388.7m | 55.9/100 | 41.5% | High | High leverage, Low liquidity, Negative revenue growth |
| SME Advance 10Lending · 2025Q4 · rating B | €305.4m | 55.7/100 | 37.4% | High | High leverage, Low liquidity, Negative revenue growth, Margin deterioration |
| Horizon Factoring 17Lending · 2025Q4 · rating BBB | €292.8m | 47.2/100 | 22.3% | Medium | High leverage, Weak interest coverage, Negative revenue growth |
| EuroMerchant 02Payments · 2025Q4 · rating BB | €267.4m | 46.6/100 | 24.7% | Medium | Low liquidity, Negative revenue growth |
| Atlas Payments 06Payments · 2025Q4 · rating A | €396.6m | 45.4/100 | 25.3% | Medium | Weak interest coverage, Negative revenue growth |
| InstantPay 15Payments · 2025Q4 · rating A | €68.2m | 42.8/100 | 2.48% | Medium | Stable financial profile |
| Horizon AM 03Asset Management · 2025Q4 · rating B | €408.5m | 41.6/100 | 49.8% | Medium | Low liquidity, Negative revenue growth, Margin deterioration |
| SME Advance 20Lending · 2025Q4 · rating BBB | €29.1m | 38.7/100 | 43.4% | Medium | Negative revenue growth, Margin deterioration |
| InstantPay 05Payments · 2025Q4 · rating B | €49.9m | 38.4/100 | 51.6% | Medium | High leverage, Weak interest coverage, Negative revenue growth, Margin deterioration |
| Atlas Payments 01Payments · 2025Q4 · rating A | €71.0m | 37.9/100 | 35.6% | Medium | Low liquidity, Negative revenue growth |
| Digital Wallet Iberia 13Payments · 2025Q4 · rating BB | €415.7m | 33.0/100 | 34.4% | Medium | High leverage, Negative revenue growth, Margin deterioration |
| IberCredit 04Banking · 2025Q4 · rating BBB | €271.3m | 33.0/100 | 0.95% | Medium | Stable financial profile |
| Castilla Insurance 07Insurance · 2025Q4 · rating A | €363.5m | 32.0/100 | 0.81% | Medium | Stable financial profile |
| Castilla Bank 18Banking · 2025Q4 · rating BB | €27.1m | 31.0/100 | 40.3% | Medium | Negative revenue growth, Margin deterioration |
| NovaBank 07Banking · 2025Q4 · rating BB | €93.2m | 30.2/100 | 21.5% | Medium | Negative revenue growth |
| Legacy Brokerage 02Asset Management · 2025Q4 · rating BBB | €159.2m | 30.0/100 | 1.40% | Medium | Stable financial profile |
| Banco Norte 16Banking · 2025Q4 · rating B | €353.2m | 29.4/100 | 16.6% | Medium | Negative revenue growth |
| PayGrid 09Payments · 2025Q4 · rating AA | €288.2m | 28.2/100 | 0.40% | Medium | Stable financial profile |
| IberCredit 09Banking · 2025Q4 · rating BBB | €356.6m | 25.0/100 | 0.81% | Medium | Stable financial profile |
| Atlas Lending 03Lending · 2025Q4 · rating AA | €303.9m | 24.4/100 | 1.57% | Low | Stable financial profile |
| Horizon AM 13Asset Management · 2025Q4 · rating B | €339.6m | 23.8/100 | 0.30% | Low | Stable financial profile |
Risk by credit rating
This adds another business lens: whether lower ratings also concentrate higher exposure and model-estimated distress probability.
| Rating | Entities | Exposure | Avg. stress score | Avg. ML probability |
|---|---|---|---|---|
| B | 13 | €2,478.6m | 39.7/100 | 33.3% |
| A | 19 | €3,917.8m | 16.7/100 | 4.00% |
| BB | 22 | €4,839.4m | 14.2/100 | 7.40% |
| BBB | 36 | €8,835.4m | 14.1/100 | 3.70% |
| AA | 9 | €2,011.4m | 13.7/100 | 0.70% |
| AAA | 1 | €125.9m | 13.1/100 | 1.20% |
How it resembles real work
The project keeps the stack simple, but the flow is realistic: clean data, create financial features, detect trend deterioration, aggregate impact, explain alerts and validate a predictive baseline.
Data engineering
Columnar PySpark transformations prepare scalable features.
Financial context
Ratios link the data to leverage, liquidity and debt-service capacity.
Business impact
Exposure at risk helps prioritise what matters economically.
Explainability
Drivers explain why an entity appears in the watchlist.
Validation
Train/test metrics avoid presenting the model as a black box.