ILOG

Contribution to AI

Constraint programming, until the early 1990s, was something you did inside a Prolog dialect. ILOG Solver, released in 1993, recast it as a C++ library — domain variables and propagation algorithms as objects, search as code callable from an ordinary program — a design Jean-François Puget described in 1994 as a C++ implementation of constraint logic programming. That packaging decision moved the technique out of research languages and into the industrial mainstream, and ILOG Schedule extended the pattern a year later by making resources and activities first-class objects over the same engine. Acquiring CPLEX Optimization in 1997 set Robert Bixby's simplex and branch-and-bound codes beside that constraint engine, so operations research and AI search were sold, documented and eventually hybridised as one toolkit instead of rival cultures; OPL, designed for the company by Pascal Van Hentenryck in 1999, gave both a single declarative modelling language. The rule products did something comparable for expert systems. Production rules and the Rete matcher outlived the collapse of the expert-systems market by being renamed business rules, kept in a repository, written in near-English by underwriters and pricing analysts, and executed inside conventional applications — the template Blaze Advisor, Drools and IBM's later decision products all worked from. It also demonstrated a route from a European public laboratory to a Nasdaq listing, selling INRIA work such as Gérard Berry's Esterel along the way.

Drafted with AI and edited by hand (claude-opus-5, reviewed 2026-08).