In a Nairobi ambulance deployment competition, the model predicted crash probability per road segment. But the real decision was: where should 6 ambulances be positioned right now?

The important part was the translation layer between prediction and action: predict risk → select highest-risk segments → cluster into geographic groups → place ambulances at centroids.

Result: In real-world ML, prediction quality and decision quality are related, but they are not the same thing.

Some categorical variables are informative but too coarse in raw form. Job_type mixes several underlying dimensions: how income is generated, how often it arrives, and how savings are managed.

I decomposed it into multiple behavioral features: income source, savings pattern, income frequency. Each derived feature captures a different financial mechanism.

Result: The variable is no longer a flat category — it becomes a behavioral representation layer separating latent mechanisms.

In a multi-entity time series problem, I spent more time on routing than on model tuning. The series were too heterogeneous: some decelerating, some accelerating, some near-zero, some ambiguous.

Before fitting anything, I built a classification layer: compute elasticity score, classify each entity into a trajectory regime, route to a different forecasting strategy per regime.

Result: When you have many entities to forecast, don't just build a better model. Build a better routing system.

I encoded six categorical variables differently based on what they represented: Helmert Sequential for day_of_week, Polynomial Ordered for hardware generations, WoE for binary splits, Target encoding for high cardinality.

Every encoding imposes an assumption. Better tabular modeling often starts with representing variables in ways that reflect their actual structure.

Result: Encoding is not preprocessing — it is part of how the model is allowed to interpret the variable.

Started from precomputed communication communities (SAS CLA results). A communication community is not necessarily a household — it can contain extended family, multiple households, and close friends.

Combined two signals: ranked communication intensity (top contacts weighted by position) and shared dominant evening cell-tower usage. Same community + top-ranked contact + shared evening tower = household tie.

Result: In applied graph analytics, the first cluster is often not the real unit you care about.

Most data science asks "what increases Y?" — leading to correlation, regression, SHAP. But some questions are different: what must already be true for high Y to even be possible?

Used Necessary Condition Analysis (NCA) to identify variables that act as minimum requirements, not average drivers. Below a threshold, the outcome simply does not appear.

Result: Some variables are not drivers — they are constraints. "Invest in X to boost" ≠ "guarantee X above threshold before anything else matters."

In a startup survival project (~50K funding rounds), the label "did not raise again" breaks if the company hasn't been observable for the full 36-month horizon. A startup founded in 2019, observed at 2021, looks like failure but simply hasn't had enough time.

Made two design decisions: eligibility window (only train on 36+ months observable history) and asymmetric holdout (mirror deployment prevalence, not balanced 50/50).

Result: Label design is often more important than algorithm choice. Once temporal contamination enters training, the model learns recency instead of risk.

In a telecom family-detection project, I needed to distinguish household-level ties from broader social communities. Added shared dominant cell-tower usage — "which cell tower shows up most in your evening CDR?" People spend ~8 hours a night in the same place.

Infrastructure data became behavioral evidence. The richest signals often hide in operational data, not in declared information.

Result: Cell tower dominance acts as a strong proxy for home location — operational traces as features.

Built an XGBoost model to predict hourly telecom data usage. Standard confidence intervals treated all predictions the same way — but peak hours had much larger errors. The business was making decisions based on false confidence.

Wrapped the model with conformal prediction: a distribution-free uncertainty quantification method. The interval width adapts to the data — wider in peak hours, narrower in off-peak.

Result: Point predictions without uncertainty are incomplete. The interval is the signal that tells you when to trust the forecast.

In restaurant rating prediction, a single engineered feature combining order frequency, recency, and menu diversity outperformed 40 raw features. Feature engineering is where domain knowledge becomes model performance.

The raw data had everything: order counts, timestamps, menu items, customer IDs. But none of those columns alone captured what actually drives a restaurant's rating.

The business rule was simple: a restaurant that gets repeat customers, recently, across different menu items, is doing something right. Encoding that logic into a single feature gave the model something it couldn't learn from the raw columns.

Result: One engineered feature outperformed 40 raw features in isolation.

In a multi-ID weekly forecasting pipeline, IDs with 26+ consecutive zeros were routed out before training. Letting the model learn zero behavior wastes capacity and distorts the loss for active series.

The model was trying to learn two completely different behaviors at once: active series with real signal, and dead series with structural zeros. The loss function couldn't distinguish between them.

The fix was a pre-processing routing step: detect dead IDs by consecutive zero count, route them to a zero-forecast rule, and train the model only on active series.

Result: Routing dead series out before training improved forecast accuracy on active series and reduced training time.

In IEEE customer support escalation prediction, the strongest signals were structural: response latency, ticket re-opening rate, message length progression, not the sentiment of the text itself.

The intuition was that escalation is a behavioral pattern, not a linguistic one. An angry customer doesn't always write angry words. But they do reopen tickets, send longer messages over time, and wait longer between responses.

Structural features captured the shape of the interaction, not what was said, but how the conversation evolved over time.

Result: Structural features outperformed raw text sentiment by a wide margin. The model predicted escalation before the customer explicitly expressed frustration.