Schneider Electric Challenge — AI-Powered Sales Opportunity Prediction & Explainability Platform

Executive Overview

Forecasting sales, and improving full stack application with machine leanirng models to display explanatory data on historical sales analyse sale segmentations and give suggestions to improve future sales

Solution Architecture Summary

Core Capability Stack

• Prediction Engine: Random Forest with optimized preprocessing and outlier filtration.
• Explainability Layer: Global + local SHAP intelligence with percent-impact analytics + interpretation LLM pipleline.
• Segmentation Engine: K-means clustering augmented with AI-authored executive briefs.
• Scenario Counterfactual Simulator: Natural language what-if analytics, LLM action interpeter + SHAP Original and Modified execution and later interpretation LLM pipeline.
• NLP Insight Generation: Gemini interprets drivers and cluster dynamics in commercial language.
• REST API: Integrated and fully documented FastAPI services.

Operating Model: How It Works

Technical Architecture

A FastAPI control plane orchestrates:

• Predictive model training & inference
• SHAP-driven explainability services
• K-means segmentation lifecycle
• Perturbation-based scenario modeling
• Gemini-assisted summarization, parsing, and impact quantification

Workflow

1. Preprocessing: LOF outlier screening and feature engineering.
2. Training: Random Forest (500 estimators), stratified sampling, F1-score validation.
3. Explainability: Global feature impact + opportunity-level rationale.
4. Clustering: 3 market-relevant behavioral segments with narrative interpretation.
5. Scenario Testing: Natural language → feature adjustments → accept/decline projections.
6. AI Interpretation: Gemini reframes analytical output into executive-grade guidance.

Key Differentiators

• Global SHAP: Tiered feature impact across the revenue pipeline.
• Local SHAP: Conversion drivers per opportunity, enabling deal-level coaching.
• Segmentation: Customer clusters with win-rate insights and executive summaries.
• Scenario Analysis: Quantified performance deltas tied to commercial levers.
• Natural Language Interaction: Interpretive commands like:
  Increase Customer Contracts by 20% → predicted uplift & attribution analysis.

Technology Stack

Business Outcomes & Wins

• Fully explainable ML workflow from ingestion to revenue-impact prediction.
• Percent-based SHAP contribution scoring for rapid sales adoption.
• Gemini-driven commercial language interpretation for stakeholder alignment.
• Scenario A/B simulation for go-to-market optimization.
• Segmentation intelligence packaged for executive consumption.
• REST API ready for integration with CRM/BI ecosystems.

Future-State Roadmap

• Real-time insight dashboard with interactive SHAP visual layers.
• Automated hyperparameter tuning with cross-validation.
• CRM-connected real-time inference architecture.
• Advanced segmentation (dynamic K, hierarchical clustering).
• Drift detection + MLOps telemetry for scalable productionization.

Built With

• data
• fastapi
• machine-learning
• python
• react