AI-Driven Pricing Platform

The Challenge

Lennar, one of America's largest homebuilders, faced a critical challenge in optimizing home prices across thousands of properties nationwide. The existing manual pricing process was:

• Time-intensive: Taking days to adjust prices across divisions
• Inconsistent: Different methodologies across regions
• Suboptimal: Leaving significant revenue on the table
• Opaque: Limited visibility into pricing rationale

The Solution

I led the architecture and development of Lennar's flagship AI-driven pricing platform, coordinating across multiple teams including QA automation, Data Engineering, Data Science, and Frontend Engineering.

System Architecture

Full-stack enterprise platform built on modern cloud-native architecture:

Pricing Optimization Engine

The core algorithm combines market data, historical sales, and AI predictions:

// Simplified pricing optimization engine
export async function optimizePrice({
  property,
  marketData,
  historicalSales,
  competitorPricing,
}: PricingInputs): Promise<OptimizedPrice> {
  // AI model integration with GPT-4o for market analysis
  const basePrice = await mlModel.predictOptimalPrice({
    features: extractFeatures(property, marketData),
    constraints: getBusinessConstraints(),
  });

  // Apply margin optimization strategy
  const optimizedPrice = applyMarginStrategy(basePrice, {
    targetMargin: property.division.targetGrossMargin,
    marketConditions: marketData.currentConditions,
    competitorPricing: competitorPricing,
  });

  return {
    recommendedPrice: optimizedPrice,
    confidence: calculateConfidence(historicalSales),
    marginImpact: calculateMarginDelta(property.currentPrice, optimizedPrice),
    rationale: generateExplanation(optimizedPrice, property),
  };
}

Key Features

1. Real-Time Price Optimization

• AI algorithms analyze market conditions, competitor pricing, and historical data
• Recommendations updated continuously as market conditions change
• Explainable AI provides clear rationale for each pricing decision
• Confidence scoring for every recommendation

2. Enterprise-Wide Integration

The platform integrates across Lennar's entire technology ecosystem:

• Seamless integration with JD Edwards, Salesforce, and RPA systems
• Unified data pipeline using DBT and Snowflake
• Real-time synchronization across all divisions
• Qlik Replicate for CDC (Change Data Capture) from legacy systems

"It operates on a Salesforce backbone, which ingests data from across the Lennar sales landscape."

— Stuart Miller, co-CEO and Executive Chairman, Lennar Q2 2025 Earnings Call

3. User-Centric Design

Built frontend UX flows ensuring seamless connection from user input to algorithmic output:

// Pricing workflow implementation
const PricingWorkflow = () => {
  const [property, setProperty] = useState<Property>();
  const [optimizedPrice, setOptimizedPrice] = useState<OptimizedPrice>();

  // Step 1: Property Selection
  const handlePropertySelect = async (propertyId: string) => {
    const propertyData = await api.getProperty(propertyId);
    setProperty(propertyData);
  };

  // Step 2: AI Price Optimization
  const handleOptimize = async () => {
    const result = await api.optimizePrice(property.id, {
      includeMarketAnalysis: true,
      generateRationale: true,
    });
    setOptimizedPrice(result);
  };

  // Step 3: Review & Adjustment
  const handleAdjust = (adjustedPrice: number) => {
    // Allow practitioners to fine-tune AI recommendations
    const updated = { ...optimizedPrice, price: adjustedPrice };
    setOptimizedPrice(updated);
  };

  // Step 4: Approval & Sync
  const handleApprove = async () => {
    // Push approved price to JDE and Salesforce
    await api.approvePrice(property.id, optimizedPrice);
    // Trigger sync across all systems
    await api.syncToERP(property.id);
  };

  return (
    <PricingDashboard
      property={property}
      optimizedPrice={optimizedPrice}
      onOptimize={handleOptimize}
      onAdjust={handleAdjust}
      onApprove={handleApprove}
    />
  );
};

UX Features:

• Intuitive React-based interface for division practitioners and regional teams
• Role-based access control with granular permissions (view, edit, approve)
• Mobile-responsive design for field access and on-the-go decision making
• Real-time collaboration with live updates across teams
• Bulk operations for adjusting thousands of properties simultaneously

Technical Deep Dive

Data Pipeline Architecture

Built a robust data pipeline handling millions of property records across multiple data sources:

Key Pipeline Features:

• Real-time CDC via Qlik Replicate for JD Edwards integration
• DBT models for data transformation and business logic
• Snowflake as the central data warehouse
• 15-minute refresh cycle for pricing data
• 99.8% data accuracy with automated validation

-- DBT transformation for margin optimization
WITH base_pricing AS (
  SELECT
    property_id,
    current_price,
    construction_cost,
    market_avg_price,
    days_on_market,
    division_id,
    community_name
  FROM {{ ref('stg_properties') }}
),
competitor_analysis AS (
  SELECT
    property_id,
    AVG(competitor_price) as avg_competitor_price,
    COUNT(*) as competitor_count
  FROM {{ ref('stg_competitor_pricing') }}
  GROUP BY property_id
),
margin_analysis AS (
  SELECT
    bp.*,
    ca.avg_competitor_price,
    ca.competitor_count,
    (bp.current_price - bp.construction_cost) / bp.current_price AS current_margin,
    PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY bp.market_avg_price)
      OVER (PARTITION BY bp.division_id) AS median_market_price,
    -- Calculate recommended price adjustment
    CASE
      WHEN bp.days_on_market > 90 THEN bp.current_price * 0.97
      WHEN ca.avg_competitor_price < bp.current_price THEN ca.avg_competitor_price * 1.02
      ELSE bp.current_price
    END as recommended_price
  FROM base_pricing bp
  LEFT JOIN competitor_analysis ca ON bp.property_id = ca.property_id
)
SELECT
  *,
  (recommended_price - current_price) as price_adjustment,
  ((recommended_price - construction_cost) / recommended_price) as projected_margin
FROM margin_analysis

Performance Optimization

Achieved sub-second response times through:

• Serverless architecture: AWS Lambda for auto-scaling
• Caching strategy: Redis for frequently accessed data
• Database optimization: Indexed queries and materialized views
• CDN distribution: CloudFront for static assets

Quality Assurance

Implemented comprehensive testing:

• Unit tests: 95% code coverage
• Integration tests: API contract testing
• E2E tests: Cypress for critical user flows
• Load testing: Handles 10,000+ concurrent users

Results & Impact

Business Impact Breakdown

Financial Results:

• $36M+ projected annual revenue uplift through optimized pricing
• 3.2% average gross margin improvement across portfolio
• ROI achieved within first quarter of deployment
• Thousands of properties optimized at scale

"We have invested heavily in the future of this high technology program, which is designed to reduce our customer acquisition cost both internal and external and manage the dynamic pricing of our homes."

— Stuart Miller, co-CEO and Executive Chairman, Lennar Q2 2025 Earnings Call

Operational Efficiency:

• 85% reduction in pricing decision time (days → minutes)
• 100% adoption across all U.S. divisions and communities
• 0 critical incidents since launch
• Automated pricing insights and margin rationale generation

"It was and still is our primary digital marketing and customer acquisition product and it has become central to our overall marketing and sales efforts."

— Stuart Miller, co-CEO and Executive Chairman, Lennar Q2 2025 Earnings Call

Technical Excellence:

• 99.9% uptime SLA consistently achieved
• <200ms p95 API response time
• 10,000+ concurrent users supported
• 95% test coverage with comprehensive CI/CD

Implementation Journey

Leadership & Collaboration

As Principal Software Engineer, I:

• Led architecture across 4 cross-functional teams (QA, Data Engineering, Data Science, Frontend)
• Coordinated execution managing system reliability and multi-team delivery
• Mentored engineers on React, AWS, and data engineering best practices
• Presented to C-suite on technical strategy and business impact
• Established standards for code quality, testing, and deployment processes
• Scaled the system from MVP to enterprise-wide adoption nationwide

Lessons Learned

What Worked Well

• Incremental rollout: Piloted in select divisions before nationwide launch, validating assumptions
• User feedback loops: Weekly sessions with practitioners shaped UX and feature prioritization
• Modular architecture: Enabled parallel development across 4 cross-functional teams
• Strong CI/CD: Azure Pipelines and GitHub Actions ensured stable, frequent deployments
• Data-first approach: Invested heavily in DBT data models and validation upfront
• Explainable AI: Transparency in pricing rationale built trust with users

Challenges Overcome

Data Integration Complexity:

• Integrated 8+ data sources (JDE, Salesforce, RPA, internal APIs)
• Built validation layers to handle inconsistent legacy data
• Implemented Qlik Replicate for real-time CDC from JD Edwards
• Created data quality monitoring with automated alerts

Cross-Team Coordination:

• Coordinated across QA, Data Engineering, Data Science, and Frontend teams
• Established clear API contracts and integration points
• Implemented feature flags for independent team deployment
• Created shared documentation and architectural decision records (ADRs)

Enterprise Scale:

• Optimized for 10x growth from initial MVP projections
• Implemented caching strategies and database indexing for performance
• Built auto-scaling infrastructure to handle variable load
• Achieved <200ms p95 latency at nationwide scale

Change Management:

• Created comprehensive training materials and video tutorials
• Conducted hands-on workshops for division practitioners
• Built in-app guidance and tooltips for complex features
• Provided 24/7 support during initial rollout phases

Executive Recognition

The platform has become a regular highlight in Lennar's quarterly earnings calls:

Stuart Miller, co-CEO and Executive Chairman, Lennar Q2 2025 Earnings Call

This project demonstrates my ability to lead complex, high-impact technical initiatives that directly drive business value at the highest executive level. The combination of technical excellence, cross-team coordination, and focus on measurable outcomes—recognized publicly by C-suite leadership as the "crown jewel" of the company's tech portfolio—exemplifies my approach to principal-level engineering.