Case studies and prototypes across analytics, data engineering, and AI — solving real problems with measurable outcomes.
Summary: Weekly revenue, conversion, and cohort retention from clickstream + orders.
Conflicting KPIs across teams.
Single source of truth dashboard.
SQL windows, semantic model, drilldowns.
Reporting time cut 6h → 30m.
Summary: Cohorts + survival curves to find churn drivers and segments.
Churn +4pp in Q2.
Prioritize save-offers.
KM curves; cohort SQL; insights deck.
12% churn reduction in pilot.
Summary: Promo/price/holiday features improved SKU forecasts.
Naive baseline under-forecast promos.
+15% MAPE lift on long tail.
Lags/rolls; Optuna; MLflow.
MAPE +21%; holding cost −8%.
Summary: Fine-tuned ResNet50; shipped inference API.
Manual QC slow.
>90% precision.
Augmentations; AMP; TTA.
92% precision; review time −70%.
Summary: Bronze–silver–gold with CDC + tests.
Nightly ETL delays.
Incremental ELT.
MergeInto, watermark, DQ checks.
24h → 30m latency.
Summary: Tuning + lineage + alerting.
Month-end cost spikes.
Cut cost, keep SLAs.
AQE, Z-Order, autoscaling.
−30% cost; 99.9% reliability.
Summary: Payments anomaly detection with explainable alerts.
Fraud review lag.
<2 min detection.
Sliding windows; IForest + rules.
FP −35%; p95 <60s.
Summary: Retrieval-augmented Q&A with guardrails & evals.
Docs scattered.
Grounded answers with citations.
Chunking, hybrid search, rerank.
Search time −80%.
Summary: Stakeholder workshops + KPI tracking.
Margin compression.
Elasticities & test plan.
Discovery → dashboards.
+3% margin in pilot.