thefd.engineer
All case studiesDistribution operations

Picking-line balancing

Which moves can the floor actually make?

Helping floor planners catch a local capacity problem before approving a stock move.

My role
AI engineering · Optimization · Operator workflow
Tools
Python · FastAPI · PostgreSQL · Mixed-integer optimization

Problem

In a distribution center, moving products between picking locations can spread demand more evenly. A move that helps one part of the line can overload another or put a product in a location that cannot handle it.

Technical decision

I kept the optimization run separate from the decision to move stock. The solver proposes a layout subject to placement constraints. An operations reviewer can inspect where work would shift and which limits the proposal would exceed.

What I built

I helped build the workflow from demand history and location data to a list of proposed stock moves. A Python worker runs the optimization. The planning interface and the floor view read the same scenario through one API, so a reviewer can trace a suggested move back to its assumptions.

Evidence

An archived scenario provided a useful check: it predicted less picking time overall, yet failed a workload limit on part of the line. The review view surfaced that failure. This verified that a reviewer could see the trade-off before acting. Realized productivity gains would need observations after the moves.

Illustration of stock locations and a proposed move that needs a capacity check.
Illustrative picking layout. The highlighted move needs a local capacity check before it can reach the movement sheet.