# Carla HQ Tabular & Time-Series AI

> Carla HQ (carlahq.eu) provides WebGPU client-side tabular and time-series AI inside Google Sheets, powered by Inria SODA's TabICLv2 and Mistral Timestral foundation models.

## Benchmark Directories & Hubs
- [Dataset Catalog & Leaderboard](https://carlahq.eu/datasets/index.md): Empirical 5-fold cross-validation across 41 open tabular & time-series benchmarks.
- [Tabular Foundation Model Comparisons](https://carlahq.eu/datasets/comparison/index.md): Head-to-head Elo ratings, win rates, and latency breakdowns.
- [Time-Series Model Comparisons](https://carlahq.eu/datasets/comparison/time-series/index.md): 100-task probabilistic forecasting benchmarks (FEV-bench).
- [Industry Use Cases & Domain Guides](https://carlahq.eu/machine-learning/index.md): Sector-specific applications across finance, healthcare, workforce, commerce, and automotive.
- [Technical Journal & Research Articles](https://carlahq.eu/blog/index.md): Deep-dives into in-browser WebGPU inference, explainability (SHAP), and synthetic prior distributions.

## Machine Learning Industry Guides
- [Machine Learning in Finance & Banking](https://carlahq.eu/machine-learning/finance-banking.md): Credit risk · customer acquisition · valuation · macroeconomic forecasting.
- [Machine Learning in Healthcare & Life Sciences](https://carlahq.eu/machine-learning/healthcare-life-sciences.md): Risk prediction · diagnosis support · patient outcomes · clinical resource planning.
- [Machine Learning in Operations & Workforce Analytics](https://carlahq.eu/machine-learning/operations-workforce.md): Demand forecasting · energy management · infrastructure telemetry · quality control.
- [Machine Learning in Commerce & Industry](https://carlahq.eu/machine-learning/commerce-industry.md): Customer behaviour · sales forecasting · inventory replenishment · pricing.
- [Machine Learning in Automotive & Mobility](https://carlahq.eu/machine-learning/automotive.md): Vehicle classification · valuation · maintenance · fleet decisions.
- [Machine Learning in Public Services & Risk](https://carlahq.eu/machine-learning/public-services-risk.md): Policy analysis · risk triage · environmental monitoring · fairness.

## Technical Articles & Research Notes
- [Demystifying SHAP: How Game Theory Explains ML](https://carlahq.eu/blog/demystifying-shap-explainable-ai.md): Shapley values, efficiency axioms, and local/global explainability.
- [Time Series Forecasting in the Browser with Timestral](https://carlahq.eu/blog/time-series-forecasting-in-the-browser-with-timestral.md): Local WebGPU execution of Mistral's 123M parameter forecasting foundation model.
- [Encoding Strings in Tabular Foundation Models](https://carlahq.eu/blog/encoding-strings-in-tabular-foundation-models.md): Handling high-cardinality text in zero-shot tabular learning.
- [Tabular Foundation Models: In-Browser Feasibility Comparison](https://carlahq.eu/blog/tabiclv2-vs-tabpfn3-vs-tabpfn26-vs-tabfm-vs-exaone-which-tabular-foundation-model-can-run-in-the-browser.md): Architectural trade-offs of TabICLv2 vs TabPFN vs EXAONE vs TabFM.
- [Synthetic Data Generation via DP-SGLD](https://carlahq.eu/blog/synthetic-data-generation-dp-sgld.md): Differential privacy and Bayesian prior synthesis.
- [Chrome Extension Manifest V3 Permissions Audit](https://carlahq.eu/blog/chrome-extension-manifest-v3-permissions-audit.md): Security analysis of client-side local WebGPU execution.
- [Google Workspace OAuth Consent Screen Approval](https://carlahq.eu/blog/oauth-consent-screen-approval.md): Zero-data-retention compliance.

## Tabular Foundation Model Comparisons
- [TabICLv2 vs TabPFN v3](https://carlahq.eu/datasets/comparison/tabiclv2-vs-tabpfn.md): Pairwise scoreboard, Elo comparison, and per-dataset deltas.
- [TabICLv2 vs EXAONE Tabular](https://carlahq.eu/datasets/comparison/tabiclv2-vs-exaone-tabular.md): Pairwise scoreboard, Elo comparison, and per-dataset deltas.
- [TabICLv2 vs Google TabFM](https://carlahq.eu/datasets/comparison/tabiclv2-vs-google-tabfm.md): Pairwise scoreboard, Elo comparison, and per-dataset deltas.
- [TabPFN v3 vs EXAONE Tabular](https://carlahq.eu/datasets/comparison/tabpfn-vs-exaone-tabular.md): Pairwise scoreboard, Elo comparison, and per-dataset deltas.
- [TabPFN v3 vs Google TabFM](https://carlahq.eu/datasets/comparison/tabpfn-vs-google-tabfm.md): Pairwise scoreboard, Elo comparison, and per-dataset deltas.
- [EXAONE Tabular vs Google TabFM](https://carlahq.eu/datasets/comparison/exaone-tabular-vs-google-tabfm.md): Pairwise scoreboard, Elo comparison, and per-dataset deltas.

## Foundation Model Profiles
- [TabICLv2 (Inria SODA)](https://carlahq.eu/datasets/models/tabiclv2.md): Architecture specs, pretraining prior, context window, and benchmark scorecards.
- [TabPFN v3 (Prior Labs)](https://carlahq.eu/datasets/models/tabpfn.md): Architecture specs, pretraining prior, context window, and benchmark scorecards.
- [EXAONE Tabular Foundation Model](https://carlahq.eu/datasets/models/exaone-tabular.md): Architecture specs, pretraining prior, context window, and benchmark scorecards.
- [Google TabFM v1.0 (Tabular Foundation Model)](https://carlahq.eu/datasets/models/tabfm.md): Architecture specs, pretraining prior, context window, and benchmark scorecards.

## Curated Benchmark Datasets
- [Abalone (regression)](https://carlahq.eu/datasets/abalone.md): Agriculture, Forestry & Fishing — 4177 rows, 9 columns. Parquet & interactive Google Sheets demo.
- [Adult (binary_classification)](https://carlahq.eu/datasets/adult.md): Economics & Public Policy — 48842 rows, 15 columns. Parquet & interactive Google Sheets demo.
- [Airfoil Self Noise (regression)](https://carlahq.eu/datasets/airfoil-self-noise.md): UCI — 1503 rows, 6 columns. Parquet & interactive Google Sheets demo.
- [Amazon Employee Access (binary_classification)](https://carlahq.eu/datasets/amazon-employee-access.md): Information Technology & Enterprise Security — 32769 rows, 10 columns. Parquet & interactive Google Sheets demo.
- [Bank Marketing (binary_classification)](https://carlahq.eu/datasets/bank-marketing.md): Finance & Banking — 45211 rows, 16 columns. Parquet & interactive Google Sheets demo.
- [BizITobs Cloud Infrastructure Telemetry (time_series_forecasting)](https://carlahq.eu/datasets/bizitobs-l2c-5t.md): Technology & Cloud Infrastructure — 31968 rows, 9 columns. Parquet & interactive Google Sheets demo.
- [Blood Transfusion Service Center (binary_classification)](https://carlahq.eu/datasets/blood-transfusion-service-center.md): Healthcare & Biomedicine — 748 rows, 5 columns. Parquet & interactive Google Sheets demo.
- [Boomlet Smart Meter Building Energy (time_series_forecasting)](https://carlahq.eu/datasets/boomlet-1282.md): Energy & Utilities — 16384 rows, 43 columns. Parquet & interactive Google Sheets demo.
- [Boomlet Smart Meter Building Energy (time_series_forecasting)](https://carlahq.eu/datasets/boomlet-1676.md): Energy & Utilities — 10463 rows, 108 columns. Parquet & interactive Google Sheets demo.
- [Boomlet Smart Meter Building Energy (time_series_forecasting)](https://carlahq.eu/datasets/boomlet-619.md): Energy & Utilities — 16384 rows, 60 columns. Parquet & interactive Google Sheets demo.
- [Breast W (binary_classification)](https://carlahq.eu/datasets/breast-w.md): Healthcare & Life Sciences — 699 rows, 10 columns. Parquet & interactive Google Sheets demo.
- [Car (multiclass_classification)](https://carlahq.eu/datasets/car.md): Automotive & Fleet Management — 1728 rows, 7 columns. Parquet & interactive Google Sheets demo.
- [Compas Two Years (binary_classification)](https://carlahq.eu/datasets/compas-two-years.md): Legal & Public Safety — 5278 rows, 14 columns. Parquet & interactive Google Sheets demo.
- [Credit G (binary_classification)](https://carlahq.eu/datasets/credit-g.md): Finance & Banking — 1000 rows, 21 columns. Parquet & interactive Google Sheets demo.
- [Diabetes (binary_classification)](https://carlahq.eu/datasets/diabetes.md): Healthcare & Biomedicine — 768 rows, 9 columns. Parquet & interactive Google Sheets demo.
- [Employee Salaries (regression)](https://carlahq.eu/datasets/employee-salaries.md): Human Resources & Workforce Analytics — 9228 rows, 11 columns. Parquet & interactive Google Sheets demo.
- [Nord Pool Day-Ahead Electricity Prices (time_series_forecasting)](https://carlahq.eu/datasets/epf-np.md): Energy & Utilities — 52416 rows, 5 columns. Parquet & interactive Google Sheets demo.
- [Electricity Transformer Temperature (time_series_forecasting)](https://carlahq.eu/datasets/ett-15t.md): Energy & Utilities — 139360 rows, 9 columns. Parquet & interactive Google Sheets demo.
- [Electricity Transformer Temperature (time_series_forecasting)](https://carlahq.eu/datasets/ett-1h.md): Energy & Utilities — 34840 rows, 9 columns. Parquet & interactive Google Sheets demo.
- [Fitness Club (binary_classification)](https://carlahq.eu/datasets/fitness-club.md): Fitness, Sports & Recreation — 1500 rows, 7 columns. Parquet & interactive Google Sheets demo.
- [Hospital Emergency Admissions (time_series_forecasting)](https://carlahq.eu/datasets/hospital-admissions-1d.md): Healthcare & Epidemiology — 13846 rows, 3 columns. Parquet & interactive Google Sheets demo.
- [Hospital Emergency Admissions (time_series_forecasting)](https://carlahq.eu/datasets/hospital-admissions-1w.md): Healthcare & Life Sciences — 1968 rows, 3 columns. Parquet & interactive Google Sheets demo.
- [House Sales (regression)](https://carlahq.eu/datasets/house-sales.md): Real Estate & Property Valuation — 21613 rows, 21 columns. Parquet & interactive Google Sheets demo.
- [Houses (regression)](https://carlahq.eu/datasets/houses.md): Real Estate & Urban Planning — 20640 rows, 9 columns. Parquet & interactive Google Sheets demo.
- [Jena Climate & Weather Telemetry (time_series_forecasting)](https://carlahq.eu/datasets/jena-weather-1h.md): Meteorology & Climate — 8784 rows, 23 columns. Parquet & interactive Google Sheets demo.
- [Monash High-Density Time Series Benchmark (time_series_forecasting)](https://carlahq.eu/datasets/m-dense-1d.md): Retail & Supply Chain — 21900 rows, 3 columns. Parquet & interactive Google Sheets demo.
- [Monks Problems 2 (binary_classification)](https://carlahq.eu/datasets/monks-problems-2.md): Data Science & Artificial Intelligence — 601 rows, 7 columns. Parquet & interactive Google Sheets demo.
- [Phoneme (binary_classification)](https://carlahq.eu/datasets/phoneme.md): Speech Processing & Acoustic Engineering — 5404 rows, 6 columns. Parquet & interactive Google Sheets demo.
- [Global Energy Forecasting Competition (GEFCom 2014) (time_series_forecasting)](https://carlahq.eu/datasets/proenfo-gfc14.md): Energy & Utilities — 17520 rows, 4 columns. Parquet & interactive Google Sheets demo.
- [Rohlik E-Grocery Daily Fulfillment Orders (time_series_forecasting)](https://carlahq.eu/datasets/rohlik-orders-1d.md): Retail & Supply Chain — 7710 rows, 17 columns. Parquet & interactive Google Sheets demo.
- [Rossmann Store Daily & Weekly Sales (time_series_forecasting)](https://carlahq.eu/datasets/rossmann-1d.md): Retail & Supply Chain — 1050330 rows, 19 columns. Parquet & interactive Google Sheets demo.
- [Rossmann Store Daily & Weekly Sales (time_series_forecasting)](https://carlahq.eu/datasets/rossmann-1w.md): Retail & Supply Chain — 148295 rows, 18 columns. Parquet & interactive Google Sheets demo.
- [Solar Photovoltaic Power Output with Weather (time_series_forecasting)](https://carlahq.eu/datasets/solar-with-weather-1h.md): Energy & Utilities — 49648 rows, 12 columns. Parquet & interactive Google Sheets demo.
- [Spambase (binary_classification)](https://carlahq.eu/datasets/spambase.md): Cybersecurity & IT Infrastructure — 4601 rows, 58 columns. Parquet & interactive Google Sheets demo.
- [Telco Customer Churn (binary_classification)](https://carlahq.eu/datasets/telco-customer-churn.md): Telecommunications & Subscription Services — 7043 rows, 21 columns. Parquet & interactive Google Sheets demo.
- [Titanic (binary_classification)](https://carlahq.eu/datasets/titanic.md): Maritime Safety & Actuarial Risk Analysis — 1309 rows, 12 columns. Parquet & interactive Google Sheets demo.
- [UCI Air Quality Chemical Gas Sensor Telemetry (time_series_forecasting)](https://carlahq.eu/datasets/uci-air-quality-1h.md): Environmental Science & Public Policy — 9357 rows, 15 columns. Parquet & interactive Google Sheets demo.
- [UK COVID-19 National Epidemiological Surveillance (time_series_forecasting)](https://carlahq.eu/datasets/uk-covid-nation-1dcumulative.md): Healthcare & Epidemiology — 2944 rows, 16 columns. Parquet & interactive Google Sheets demo.
- [US Macroeconomic Consumption & Disposable Income (time_series_forecasting)](https://carlahq.eu/datasets/us-consumption-1y.md): Finance & Macroeconomics — 1984 rows, 3 columns. Parquet & interactive Google Sheets demo.
- [Vehicle (multiclass_classification)](https://carlahq.eu/datasets/vehicle.md): Automotive Engineering & Computer Vision — 846 rows, 19 columns. Parquet & interactive Google Sheets demo.
- [Global Sovereign Carbon Dioxide (CO2) Emissions (time_series_forecasting)](https://carlahq.eu/datasets/world-co2-emissions.md): Environmental Science & Macroeconomics — 11460 rows, 3 columns. Parquet & interactive Google Sheets demo.