Application Portfolio Rationalization Engine
A static-code-analysis system that maps a sprawling enterprise application portfolio using knowledge graphs and graph neural networks, surfacing consolidation candidates data-driven rather than by intuition.
A large enterprise carried an unwieldy application portfolio with poor visibility into overlap, dependencies, and redundancy. Consolidation decisions were guesswork, and maintaining the status quo tied up 40–50 FTEs in redundant work.
Built a knowledge-graph-based rationalization tool that analyzes source code statically and applies graph neural networks to surface dependencies, redundancy clusters, and consolidation candidates across the full portfolio.
~$1M in annual savings across a 40–50 FTE portfolio by enabling data-driven consolidation decisions.
Most portfolio rationalization efforts happen in spreadsheets fed by surveys — neither the data nor the analysis is trustworthy, so consolidation decisions stall in politics. Replacing that with code-derived evidence fundamentally changed the conversation: the graph made dependencies visible and made the cost of the status quo legible.
The graph neural network layer was the non-obvious choice. Static analysis gives you a dependency graph; GNNs let you reason about it at a structural level — identifying clusters of applications that are tightly coupled but organizationally separate, or applications that appear redundant but share no code path. That structural reasoning is what produced actionable consolidation candidates rather than a raw dependency map.
This ran as an INFINITY Labs intrapreneur project, which meant taking it from research hypothesis through validated PoC to a result the business could act on — without a pre-committed product budget.