AI Resource Planning Platform

The Problem

The client, a commercial restoration contractor, handles disaster recovery for commercial properties. When a building floods or catches fire, they mobilize crews to begin restoration work.

Their challenge: matching the right employees to the right projects at the right time.

They had:

• 44 employees with different certifications, skills, and availability
• 84+ active projects at any given time
• Constantly shifting schedules as jobs completed and new ones started
• Manual tracking in spreadsheets and a clunky Procore setup

Project managers spent hours each week figuring out who could work what job. They'd miss opportunities because they didn't realize a certified technician was available. They'd overbook people or leave skilled workers underutilized.

The Solution

I built an AI-powered platform that automates both opportunity discovery and resource matching.

Opportunity Discovery

The platform continuously scans for new project opportunities:

Web Scraping

Automated crawlers monitor relevant sources: public bid boards, insurance claim databases, industry news. When a potential restoration project surfaces, the system captures it.

LLM Classification

Raw scraped data is messy. The AI classifies opportunities by:

• Project type (water damage, fire restoration, mold remediation)
• Estimated scope and duration
• Required certifications
• Geographic location
• Timeline urgency

High-quality opportunities surface automatically instead of relying on manual research.

Employee-Project Matching

The core intelligence: a bulk recommendation agent that matches employees against open assignments.

Matching Criteria:

• Skills: Does the employee have the required capabilities?
• Certifications: Are they certified for this type of work?
• Availability: Are they free during the project window?
• Location: How far is the job from their home base?
• Experience: Have they worked similar projects before?

How It Works:

The agent analyzes all 44 employees against all open assignments, generating ranked recommendations. Project managers see a prioritized list of who should work what, with explanations for each recommendation.

Daily Alerts

The system runs continuously, surfacing high-fit opportunities as they emerge. Daily alerts notify managers when:

• A new opportunity matches available skilled workers
• An employee becomes available who's perfect for an active project
• A project is understaffed relative to scope
• Certification expirations will affect upcoming assignments

Technical Implementation

Hybrid Backend:

• Express handles real-time API endpoints
• Flask manages batch processing and ML workloads
• Azure SQL stores structured project and employee data

Scraping Infrastructure:

• Scheduled crawlers for target sites
• Change detection to avoid duplicate processing
• Content extraction pipelines for different source formats

Recommendation Engine:

• Feature engineering from employee and project attributes
• LLM-powered reasoning for complex matching decisions
• Confidence scoring for recommendation quality

Results

The platform replaced the client's Procore deployment entirely. They now use it weekly for all resource planning decisions.

Key outcomes:

• 44 employees efficiently matched against projects
• 84 projects actively tracked with full visibility
• Daily alerts surface opportunities before competitors
• Weekly planning reduced from hours to minutes

Delivered in 12 weeks from initial scoping to production deployment.