As we enter the final weeks of 2025, a natural time to remember all we are grateful for, we have something truly special to celebrate – 40 years of AMPL.
Since its inception in the fall 1985, AMPL has grown from an ambitious idea into a globally recognized system for mathematical modeling and optimization. Over four decades, it has evolved into a tool for educators to teach optimization, for researchers to study optimization, and, most importantly, for OR and Analytics professionals to transform complex operational and business challenges into clear, data-driven decisions.
This milestone wouldn’t have been possible without the vision of AMPL’s creators, the dedication of its team, and the trust of the global community that has used AMPL to solve problems across industries.
Image of AMPL team at INFORMS Analytics+ 2025
AMPL’s story begins in 1975, when Bob Fourer and David Gay both worked at the NBER Computer Research Center for Economics and Management Science in Cambridge, Massachusetts.
Bob Fourer: “I first encountered mathematical optimization at NBER, where I worked for two years in the mid-1970s. I joined a small team led by William Orchard-Hays, one of the pioneers of computational optimization, whose goal was to develop a new linear programming system.”
David Gay: “In the middle 1970s I was working on nonlinear data fitting at NBER, when I met Bob Fourer, who had recently graduated from MIT.”
In 1975, Bob Fourer and Michael J. Harrison began collaborating at NBER on a technical report that proposed “A Modern Approach to Computer Systems for Linear Programming”. It described an optimization modeling language and proposed embedding the language in a system that would encompass data preparation, solvers, and dissemination of results.
Subsequently David Gay joined AT&T Bell Laboratories’ legendary Computing Science Research Center in Murray Hill, New Jersey – source of the Unix operating system, the C programming language, and many other computer languages and tools. He was responsible for a broad variety of contributions to optimization projects there.
Bob Fourer went to Stanford for his PhD in Operations Research, working in the Systems Optimization Laboratory founded by the legendary George Dantzig, and then joined the Industrial Engineering and Management Science faculty at Northwestern University. In 1983 he published “Modeling Language versus Matrix Generators for Linear Programming” which made the case for optimization modeling languages as an essential aspect of optimization practice.
The AMPL team came together a few years later.
David Gay: “I saw Bob at a meeting and learned he was coming up for a sabbatical. I talked with my management, and they arranged for Bob to spend his sabbatical at Murray Hill. While he was there, he, Brian Kernighan, and I had many discussions and invented AMPL.”
Brian Kernighan recalled: “Bob and I met at Bell Labs in 1985, where he frequently came to work with Dave.”
Here’s a group photo from that time:
Bell Labs, 1985. Provided by Brian Kernighan.
Brian Kernighan: “I look at that picture occasionally – literally, it’s on my wall – and I still remember who a lot of those people were. With some of them, we still meet regularly.”
David Gay: “Brian Kernighan was one of the department heads when I joined Bell Labs. He was well known. He had written some very good books, sometimes with coauthors, such as the famous Kernighan and Ritchie book on programming in C”.
Bob Fourer: “An appendix to my 1983 paper listed 7 optimization modeling languages that were already available for use. But with rapid advances in computing, and with the advantage of Brian’s experience creating languages and Dave’s expertise in optimization, we believed we could build something better. We contributed our different perspectives to putting the AMPL together.”
Brian Kernighan: “It was an interesting language design project: could you design a language that was easier for expressing algebraic concepts? The project served as a perfect “sandbox” to experiment with different ideas and C++, which was a brand-new and exciting language at the time.”
Within a couple of months, a primitive but functional linear-modeling system was running. It emphasized clean, math-like syntax, fast interactive processing, and support for multiple linear solvers.
David Gay explains: “Brian wrote the first AMPL translator. To make giving data to it easier, I wrote a preprocessor that handled data sections. Later, to support extensions, I rewrote much of the translator and incorporated the data preprocessor.”
Bob Fourer adds: “Dave wrote the AMPL Solver Library (ASL) which provided standard operations used in connecting to any solver. We used that library to build interfaces for particular solvers.”
Brian Kernighan: “The first version of the AMPL translator was probably only two or three thousand lines of C++. I still have a copy of it.”
AMPL’s readable, modular, solver-independent design remains a key aspect of its appeal today.
Bob Fourer: “Richard Stone wrote some of the first industrial AMPL applications, at AT&T Bell Laboratories and then Northwest Airlines.”
Throughout the late 1980s, 1990s, and 2000s, AMPL expanded to support:
During this time, AMPL became the modeling system of choice for increasing numbers of instructors, researchers, and professionals who valued its clarity, flexibility, and solver independence.
A major turning point came in 2000 with the burst of the dot-com bubble.
David Gay recalls: “The ‘dot com bubble’ burst. … Lucent Technologies, which had become the parent company of Bell Labs, fell on hard times. The AMPL creators decided to continue development through their own company. After two years of paperwork and negotiations, we got permission from Lucent Technologies to take over the technology.”
This led to a second milestone in 2003 when the three creators of AMPL officially founded AMPL Optimization, Inc.
INFORMS International Meeting, Hong Kong, June 2006
A third critical milestone was reached in 2010, marking the company’s transition to a new phase of growth.
David Gay: “Our friend Sanjay Saigal introduced us to Bill Wells. We all met at my house in Albuquerque and convinced Bill to become our CEO. That’s when the company began to grow”.
Bob Fourer: “There was a slow buildup until we started the AMPL company. It started to grow after 2010 when we brought Bill on as the CEO to join us. Then people like Filipe joined, who brought knowledge about where computing was going. Building APIs (particularly the Python API) was a big milestone.”
Milestones Along the Way
Brian Kernighan: “The one thing that I tried early on that never really came to fruition until fairly recently was to have AMPL be a, basically, a giant subroutine, so that you could call it from something else.”
Today, AMPL sits at the intersection of optimization, data science, and intelligent computation. It has grown far beyond a modeling language into a full ecosystem that includes:
APIs allowing you to connect to other programming languages
Seamless Python integration through amplpy and data connectors
Containerized and cloud-ready deployment
Tools that plug optimization into modern data-science pipelines
AI assistance from amplbot and other AI tools
Bob Fourer notes: “The history of AMPL’s development continues to the present. The speed of developments has increased more recently with new environments for working with optimization, new APIs, and new ways of describing things”.
Bill Wells adds: “Between 2017–2023 the main progress was on the development side. Last year we expanded our marketing, sales, and service divisions — and this progress is a milestone in itself.”
Filipe Brandão (Head of Development) highlights: “The key developments were APIs, support for containers, and dynamic licensing. Dating to as early as 2017, these were innovations in optimization software that enabled the transition to the cloud.”
Brian Kernighan: “Too often, researchers invest time in C++ or open-source workarounds simply because they’re unaware of AMPL. Our mission is to demonstrate that for the optimization embedded in their work, AMPL offers a simpler and more direct path.”
As the next decade unfolds, optimization will grow even more deeply connected with AI, machine learning, and new computational paradigms.
Across CPUs, GPUs, and emerging quantum architectures, one principle remains: AMPL will continue enabling clear, expressive, and adaptable modeling for the challenges of tomorrow.
Bob Fourer: “Writing mathematical models is exactly the kind of task people now expect chatbots to handle, and they do it well. Today’s AI systems can not only write models, but also generate the surrounding software needed to turn those models into full applications. Our own experiments show that chatbots already know how to write good AMPL models because they were trained on publicly available models and examples.”
Bob also highlights the deeper transformation underway: “The more transformative impact comes from combining machine learning and optimization. If you solve a model for many different inputs, you can collect those input–output pairs and train a neural network to approximate the optimization’s results. The neural network will be much faster and can deliver near-optimal answers instantly, essential for large-scale, real-time applications. This kind of ‘learning to optimize’ is just one example of how AI is reshaping optimization. The question now is how far this integration will go?”
AMPL’s story has always been a story of people. We extend our deep gratitude to everyone who has shaped its journey:
Your engagement keeps AMPL at the forefront of the optimization industry.
As we mark forty years of AMPL, we extend our warm thanks to everyone who has been part of our story.
Your trust, creativity, and curiosity have kept AMPL vibrant through decades of technological change, and will continue carrying it into the next era of modeling and optimization.
Here’s to 40 years of innovation and to many more ideas, models, and breakthroughs ahead.
Happy Modeling,
The AMPL Team
1. A video interview with co-founder Bob Fourer, where he discusses the early days of AMPL: https://dev.ampl.com/ampl/videos/interviews.html
2. A personal history by David Gay, detailing the project’s beginnings: https://ampl.com/about/ampl-history-david-gay/
3. The homepage of Brian Kernighan: https://www.cs.princeton.edu/~bwk/
4. About us: https://ampl.com/about/
Prepared by: Mikhail Riabtsev
in collaboration with Bob Fourer, David Gay, Brian Kernighan, Bill Wells, and Filipe Brandao.
As we enter the final part of the year, a natural time to remember all we are grateful for, we at AMPL have something truly special to celebrate – 40 years of AMPL.
Since its inception in the fall 1985, AMPL has grown from an ambitious idea into a globally recognized system for mathematical modeling and optimization. Over four decades, it has evolved into a tool for educators to teach optimization, for researchers to study optimization, and, most importantly, for OR and Analytics professionals to transform complex operational and business challenges into clear, data-driven decisions.
This milestone wouldn’t have been possible without the vision of AMPL’s creators, the dedication of its team, and the trust of the global community that has used AMPL to solve problems across industries. This article is a tribute to you all.
AMPL’s story begins in 1975, when Robert Fourer and David Gay both worked at the NBER Computer Research Center for Economics and Management Science in Cambridge, Massachusetts.
Bob Fourer: “I first encountered mathematical optimization at NBER, where I worked for two years in the mid-1970s. I joined a small team led by William Orchard-Hays, one of the pioneers of computational optimization, whose goal was to develop a new linear programming system.”
Dave Gay: “In the middle 1970s I was working on nonlinear data fitting at NBER, when I met Bob Fourer, who had recently graduated from MIT.”
In 1975, Bob Fourer and Michael J. Harrison began collaborating at NBER on a technical report that proposed “A Modern Approach to Computer Systems for Linear Programming”. It described an optimization modeling language and proposed embedding the language in a system that would encompass data preparation, solvers, and dissemination of results.
Subsequently Dave Gay joined AT&T Bell Laboratories’ legendary Computing Science Research Center in Murray Hill, New Jersey – source of the Unix operating system, the C programming language, and many other computer languages and tools. He was responsible for a broad variety of contributions to optimization projects there.
Bob Fourer went to Stanford for his PhD in Operations Research, working in the Systems Optimization Laboratory founded by the legendary George Dantzig, and then joined the Industrial Engineering and Management Science faculty at Northwestern University. In 1983 he published “Modeling Language versus Matrix Generators for Linear Programming” which made the case for optimization modeling languages as an essential aspect of optimization practice.
The AMPL team came together a few years later.
Dave Gay: “I saw Bob at a meeting and learned he was coming up for a sabbatical. I talked with my management, and they arranged for Bob to spend his sabbatical at Murray Hill. While he was there, he, Brian Kernighan, and I had many discussions and invented AMPL.”
Brian Kernighan recalled: “Bob and I met at Bell Labs in 1985, where he frequently came to work with Dave.”
Here’s a group photo from that time:
Bell Labs, 1985. Provided by Brian Kernighan.
Brian Kernighan: “I look at that picture occasionally – literally, it’s on my wall – and I still remember who a lot of those people were. With some of them, we still meet regularly.”
Dave Gay: “Brian Kernighan was one of the department heads when I joined Bell Labs. He was well known. He had written some very good books, sometimes with coauthors, such as the famous Kernighan and Ritchie book on programming in C”.
Bob Fourer: “An appendix to my 1983 paper listed 7 optimization modeling languages that were already available for use. But with rapid advances in computing, and with the advantage of Brian’s experience creating languages and Dave’s expertise in optimization, we believed we could build something better. We contributed our different perspectives to putting the AMPL together.”
Brian Kernighan: “It was an interesting language design project: could you design a language that was easier for expressing algebraic concepts? The project served as a perfect “sandbox” to experiment with different ideas and C++, which was a brand-new and exciting language at the time.”
Within a couple of months, a primitive but functional linear-modeling system was running. It emphasized:
Dave Gay explains: “Brian wrote the first AMPL translator. To make giving data to it easier, I wrote a preprocessor that handled data sections. Later, to support extensions, I rewrote much of the translator and incorporated the data preprocessor.”
Bob Fourer adds: “Dave wrote the AMPL Solver Library (ASL) which provided standard operations used in connecting to any solver. We used that library to build interfaces for particular solvers.”
Brian Kernighan: “The first version of the AMPL translator was probably only two or three thousand lines of C++. I still have a copy of it.”
AMPL’s readable, modular, solver-independent design remains a key aspect of its appeal today.
Robert Fourer: “Richard Stone wrote some of the first industrial AMPL applications, at AT&T Bell Laboratories and then Northwest Airlines.”
Throughout the late 1980s, 1990s, and 2000s, AMPL expanded to support:
During this time, AMPL became the modeling system of choice for increasing numbers of instructors, researchers, and professionals who valued its clarity, flexibility, and solver independence.
A major turning point came in 2000 with the burst of the dot-com bubble.
David Gay recalls: “The ‘dot com bubble’ burst. … Lucent Technologies, which had become the parent company of Bell Labs, fell on hard times. The AMPL creators decided to continue development through their own company. After two years of paperwork and negotiations, we got permission from Lucent Technologies to take over the technology.”
This led to a second milestone in 2003 when the three creators of AMPL officially founded AMPL Optimization, Inc.
INFORMS International Meeting, Hong Kong, June 2006
A third critical milestone was reached in 2010, marking the company’s transition to a new phase of growth.
David Gay: “Our friend Sanjay Saigal introduced us to Bill Wells. We all met at my house in Albuquerque and convinced Bill to become our CEO. That’s when the company began to grow”.
Robert Fourer: “There was a slow buildup until we started the AMPL company. It started to grow after 2010 when we brought Bill on as the CEO to join us. Then people like Filipe joined, who brought knowledge about where computing was going. Building APIs (particularly the Python API) was a big milestone.”
Milestones Along the Way
Brian Kernighan: “The one thing that I tried early on that never really came to fruition until fairly recently was to have AMPL be a, basically, a giant subroutine, so that you could call it from something else.”
Python Ecosystem (amplpy):
and more...
File-Based Data:
and more...
Database:
and more...
ERP & BI Systems:
and more...
Machine Learning:
- ChatGPT
- DeepSeek
- Claude
Today, AMPL sits at the intersection of optimization, data science, and intelligent computation. It has grown far beyond a modeling language into a full ecosystem that includes:
Robert Fourer notes: “The history of AMPL’s development continues to the present. The speed of developments has increased more recently with new environments for working with optimization, new APIs, and new ways of describing things”.
Bill Wells adds: “Between 2017–2023 the main progress was on the development side. Last year we expanded our marketing, sales, and service divisions — and this progress is a milestone in itself.”
Filipe Brandão (Head of Development) highlights: “The key developments were APIs, support for containers, and dynamic licensing. Dating to as early as 2017, these were innovations in optimization software that enabled the transition to the cloud.”
Brian Kernighan: “Too often, researchers invest time in C++ or open-source workarounds simply because they’re unaware of AMPL. Our mission is to demonstrate that for the optimization embedded in their work, AMPL offers a simpler and more direct path.”
As the next decade unfolds, optimization will grow even more deeply connected with AI, machine learning, and new computational paradigms.
Across CPUs, GPUs, and emerging quantum architectures, one principle remains: AMPL will continue enabling clear, expressive, and adaptable modeling for the challenges of tomorrow.
Robert Fourer: “Writing mathematical models is exactly the kind of task people now expect chatbots to handle, and they do it well. Today’s AI systems can not only write models, but also generate the surrounding software needed to turn those models into full applications. Our own experiments show that chatbots already know how to write good AMPL models because they were trained on publicly available models and examples.”
Robert also highlights the deeper transformation underway: “The more transformative impact comes from combining machine learning and optimization. If you solve a model for many different inputs, you can collect those input–output pairs and train a neural network to approximate the optimization’s results. The neural network will be much faster and can deliver near-optimal answers instantly, essential for large-scale, real-time applications. This kind of ‘learning to optimize’ is just one example of how AI is reshaping optimization. The question now is how far this integration will go?”
AMPL’s story has always been a story of people. We extend our deep gratitude to everyone who has shaped its journey:
Your engagement keeps AMPL at the forefront of the optimization industry.
As we mark forty years of AMPL, we extend our warm thanks to everyone who has been part of our story.
Your trust, creativity, and curiosity have kept AMPL vibrant through decades of technological change, and will continue carrying it into the next era of modeling and optimization.
Here’s to 40 years of innovation and to many more ideas, models, and breakthroughs ahead.
With gratitude,
The AMPL Team
1. A video interview with co-founder Robert Fourer, where he discusses the early days of AMPL: https://dev.ampl.com/ampl/videos/interviews.html
2. A personal history by David Gay, detailing the project’s beginnings: https://ampl.com/about/ampl-history-david-gay/
3. The homepage of Brian Kernighan: https://www.cs.princeton.edu/~bwk/
4. About us: https://ampl.com/about/
Marketing Team