ML & Forecasting · Sustainability
Agricultural CO₂ Emissions Modeling
Agricultural emissions come from interdependent structural and environmental factors that are hard to model jointly.
- My role
- Data scientist intern — modeling, validation, and the scenario API.
- When
- Jun – Sep 2024
68%
variance from livestock + fertilizer
7.9%
forecast MAPE (ARDL baseline 12.4%)
26
cantons × sub-sectors modeled
API
for NGO scenario exploration
Approach
Predictive models combining agricultural, demographic, and climate features, benchmarked against econometric baselines for robustness and interpretability.
Results
Identified livestock and fertilizer as the primary emission drivers (68% of variance) across Swiss agricultural sub-sectors. XGBoost reached 7.9% MAPE versus 12.4% for the ARDL baseline, and the scenario-ready forecasts were packaged into a Python API for policy exploration at gaea21.
Architecture
Technical highlights
- 01Integrated heterogeneous climate and structural datasets
- 02Compared econometric and ML approaches for robustness and interpretability
- 03Outputs framed for policy interpretation (gaea21 NGO use case)
Stack
Related case studies
Quant Risk · MLOps
EVT Pre-Trade Limits Recalibration
94%
precision at a 1.3% false-positive rate
AI Engineering · Deep Learning
GPT from Scratch
10.8M
parameter decoder-only transformer, built from zero
ML & Forecasting · Retail
AI Pricing & Promotion Decision Engine
+6.8%
simulated margin vs a rule-based promo calendar