Fix README: spread screenshots one-at-a-time, fix broken image paths, remove duplicate architecture

README.md

@@ -34,6 +34,8 @@ If you are reviewing this project for a **Senior Data Scientist**, **Machine Lea
 * **Predictive & Segment ML Engines:** Built a **LightGBM time-series revenue forecasting model** (with residual analysis) and an **RFM K-Means Clustering model** (validated via Silhouette & Davies-Bouldin scores with PCA 2D projections).
 * **Automated AI Executive Insights:** Includes an automated C-suite report generator synthesizing revenue health, segment risks, and forecast projections into actionable strategic recommendations.
 
+![Executive Dashboard](https://raw.githubusercontent.com/gautamhardik/Analytica/main/DOCS/screenshots/01_executive_dashboard.png)
+
 ---
 
 ## 🎥 Full System Walkthrough & Demo Video
@@ -44,52 +46,6 @@ https://github.com/user-attachments/assets/e9e9354d-bc9d-4434-b829-712ba0c8069a
 
 ---
 
-## 📸 Executive Platform Showcase (Live Dashboard Screenshots)
-
-### 1. C-Suite Executive Overview & KPI Dashboard
-*Real-time executive cockpit displaying revenue metrics, order velocity, customer acquisition, and dynamic multi-dimensional filters.*
-![Executive Dashboard](DOCS/screenshots/01_executive_dashboard.png)
-
-<br/>
-
-### 2. Machine Learning Customer Segmentation (RFM + PCA K-Means)
-*Unsupervised ML model segmenting 100k+ customers into behavioral personas with 2D PCA cluster visualization and metrics.*
-![Customer Segmentation](DOCS/screenshots/06_ml_segmentation.png)
-
-<br/>
-
-### 3. LightGBM Time-Series Revenue Forecasting Engine
-*Predictive time-series model forecasting daily revenue trends with historical accuracy comparison and residual error distributions.*
-![Revenue Forecasting](DOCS/screenshots/07_revenue_forecasting.png)
-
-<br/>
-
-### 4. Automated AI Business Summary & C-Suite Report Generator
-*Data synthesis engine combining financial health, customer churn risks, and category growth opportunities into dynamic markdown reports.*
-![AI Executive Summary](DOCS/screenshots/08_ai_executive_summary.png)
-
-<br/>
-
-<details>
-<summary><b>🔍 View Additional Workspace Screenshots (Sales, Customers, Products, Geography, Reports)</b></summary>
-
-#### Sales & Revenue Analytics Workspace
-![Sales Analytics](DOCS/screenshots/02_sales_analytics.png)
-
-#### Customer Health & Retention Analytics
-![Customer Health](DOCS/screenshots/03_customer_health.png)
-
-#### Product & Category Performance
-![Product Performance](DOCS/screenshots/04_product_performance.png)
-
-#### Geographic Revenue Distribution
-![Geographic Analytics](DOCS/screenshots/05_geographic_analytics.png)
-
-#### Enterprise Data Reports Explorer
-![Data Reports Explorer](DOCS/screenshots/09_data_reports_explorer.png)
-
-</details>
-
 ---
 
 ## 🏗️ System Architecture
@@ -121,6 +77,8 @@ flowchart TD
 ```
 | Deployment | Docker |
 
+![AI Executive Summary](https://raw.githubusercontent.com/gautamhardik/Analytica/main/DOCS/screenshots/08_ai_executive_summary.png)
+
 ---
 
 ## What this project demonstrates
@@ -134,37 +92,6 @@ flowchart TD
 
 ---
 
-## Architecture
->>>>>>> 88deff280f44abd7f8785c4f8e41bf841b4bc0eb
-
-```mermaid
-flowchart TD
-    subgraph Data Layer
-        A[Olist Raw CSVs] -->|Polars ETL| B[(MySQL Star Schema)]
-        B --> C[reporting_sales_summary]
-        B --> D[reporting_customer_summary]
-        B --> E[reporting_state_summary]
-    end
-
-    subgraph ML Layer
-        F[LightGBM Forecast Model] --> H[FastAPI Backend]
-        G[RFM K-Means Segmentation] --> H
-        B -->|async SQLAlchemy| H
-    end
-
-    subgraph API Layer
-        H[FastAPI Async Backend] <--> I[Redis / TTL Cache]
-    end
-
-    subgraph Presentation Layer
-        J[Next.js Dashboard] <-->|TanStack Query| H
-        J --> K[AI Executive Summary]
-        J --> L[Filter Bar]
-    end
-```
-
----
-
 ## Engineering decisions (and why I made them)
 
 **Why a star schema over a snowflake schema:** the dashboard's query patterns are almost entirely single-join aggregations (revenue by state, revenue by month, revenue by segment). A star schema keeps those joins shallow and avoids the multi-hop joins a snowflake design would force for the same queries.
@@ -173,8 +100,12 @@ flowchart TD
 
 **Why LightGBM over ARIMA/Prophet for forecasting:** the revenue series has irregular seasonality and several categorical drivers (state, category) that a pure time-series model can't easily incorporate. A gradient-boosted tree model lets me treat forecasting as regression with time-based and categorical features, at the cost of losing some classical time-series interpretability — a tradeoff I'm explicit about in the model docs.
 
+![Revenue Forecasting](https://raw.githubusercontent.com/gautamhardik/Analytica/main/DOCS/screenshots/07_revenue_forecasting.png)
+
 **Why K-Means over hierarchical/DBSCAN for segmentation:** RFM features are low-dimensional and roughly convex per cluster after scaling, which is exactly where K-Means performs well and stays interpretable for a business audience. I validated cluster count with silhouette score and Davies-Bouldin index rather than picking K arbitrarily.
 
+![Customer Segmentation](https://raw.githubusercontent.com/gautamhardik/Analytica/main/DOCS/screenshots/06_ml_segmentation.png)
+
 **Why a caching layer at all:** several endpoints aggregate across the full fact table on every request. Rather than optimize every query indefinitely, I cached the aggregation layer — the more common and more defensible fix in a real system with a mixed read/write pattern.
 
 ---