01
Feature preparation
Historical plant demand, supplier realization, coal quality, shipment lead time, and contract values are pulled from governed warehouse models.
Back to DATATEAM projects
P2EP AI Coal Planning
A machine learning and optimization platform for forecasting demand, modeling coal supply constraints, and recommending allocation plans across PLN EPI operations.
Problems we had to solve
- Coal planning depended on analyst-maintained spreadsheets that were difficult to audit, version, and rerun under changing assumptions.
- Planning decisions had to balance demand, coal quality, plant constraints, supplier commitments, delivery lead time, and cost.
- Scenario comparison was slow because each revision required manual spreadsheet changes and repeated reconciliation.
What we implemented
- Built a planning pipeline that reads governed warehouse tables, prepares planning features, and produces demand and supply recommendations.
- Implemented optimization logic to evaluate allocation options against operational constraints and objective trade-offs.
- Created scenario outputs that planners can compare, inspect, and rerun with updated assumptions instead of rebuilding spreadsheets.
Tools used
Stack selected for production data work, not a demo.
AWSAWS ECS/Fargate
Terraform
Git
Python
scikit-learn
Stable-Baselines3 (PPO/DQN)
Snowflake
FastAPI
SvelteArchitecture
How the project is structured technically.
02
Forecasting layer
Demand and operational input forecasts are generated as planning features, then written back for review and downstream optimization.
03
Optimization layer
A constrained solver evaluates allocation candidates and produces recommendations based on cost, availability, quality, and fulfillment requirements.
Work showcase
Sanitized project screens
AI orchestration overview
Purpose-built design, engineering, and infrastructure architecture — crafted for clarity, scale, and production reliability.
Related PLN EPI work
Foundation layer
Data Warehouse
A governed Snowflake warehouse that standardizes raw operational data into bronze, silver, and gold layers for planning, operations, BI, and AI workloads.
Open detailWorkflow layer
Integrated Coal Operations
An operational platform that connects coal contracts, shipments, plant stock, billing, and monitoring workflows into one data-backed system.
Open detailAccess layer
LLM Analytics Agent
A natural-language analytics agent connected to governed operational data, allowing users to ask business questions while keeping answers grounded in approved warehouse models.
Open detail