12 projects with interactive demos. Click, interact, and see model outputs — predictions, visualizations, and scores — not code.
Organized by learning paradigm for teams evaluating data science depth.
Outcome prediction, risk scoring, forecasting — models trained on labeled data to drive measurable business decisions.
Toggle customer features and watch churn probability update in real time with confusion matrix, ROC curve, and feature importance.
89%
Accuracy
87%
Precision
91%
Recall
0.89
F1 Score
Submit a sample transaction and get a fraud risk score with SHAP waterfall explanation and precision-recall tradeoff.
0.91
PR-AUC
-34%
False Positives
-41%
Review Time
<120ms
Inference Latency
Mini form: input applicant details → instant approval probability, risk tier, and fairness metrics across demographics.
0.94
AUC-ROC
72%
Approval Rate
<3%
Fairness Gap
<50ms
Latency
Anomaly detection, segmentation, pattern discovery — extracting structure when labels are limited or unavailable.
Interactive 3D cluster visualization — hover over points to see customer profiles with elbow plots and silhouette scores.
7
Segments
0.62
Silhouette
+19%
Campaign CTR
+14%
CAC Efficiency
Live-style feed showing normal vs. anomalous data points flagged in real time with threshold tuning slider.
94%
Detection Accuracy
<100ms
Latency
10K+
Sensors
5M+
Events/Day
Visual map of whale wallet movements with behavioral clustering, temporal heatmaps, and outlier identification.
78%
Forecast Accuracy
1000+
Data Points/Min
11+
API Integrations
<200ms
Response Time
Neural networks, transformers, RAG pipelines, agent orchestration — advanced architectures for language, sequence, and multi-modal tasks.
Full-stack ML platform with FRED integration (800K+ datasets), GPT-4 AI assistant, and Docker orchestration.
15+
Docker Services
800K+
Datasets
25+
API Endpoints
6
Data Sources
Type text → real-time sentiment + intent classification with attention heatmap and confidence distribution.
92%
Accuracy
1000/min
Processing
5
Languages
3
Classes
Enter a research question → watch agents search, reason, and synthesize an answer with cited sources.
94%
Answer Quality
5-12
Sources Cited
3-7
Agent Steps
<8s
Response Time
Chat over documents — see retrieved chunks, knowledge graph traversal path, and answers with inline citations.
0.89
Retrieval MRR
91%
Answer Accuracy
50K+
KG Entities
<3s
Latency
Select a ticker to see predicted vs. actual prices with confidence intervals and interactive candlestick charts.
$2.01
RMSE
$1.54
MAE
3.2%
MAPE
0.964
R²
Upload any dataset — CSV, PDF, images, JSON. The system auto-detects the task, trains a model in real time with a live loss curve, and lets you query it instantly with built-in SHAP/attention explainability.
5
Data Modalities
96%
Auto-Detect Accuracy
<30s
Train → Inference
100%
Explainability
I can walk through model design choices, tradeoffs, deployment architecture, and the measured impact from each project.