Grant Ingersoll and the art of finding what matters

Grant Ingersoll had helped build a company around finding things. Then he encountered a problem that no search engine could solve for him: too much of the company’s work kept finding its way back to Grant Ingersoll. A founder knows the code. A founder knows the customer. Before long, the founder knows exactly why nothing can happen until the founder has a spare minute. The spare minute develops a waiting list.

At Lucidworks, he had been handling both CTO and engineering leadership responsibilities. In 2016, he described the difficult business of handing part of that work to someone else. His friend Chris Bouton had pressed a useful question: did Ingersoll want his ideas carried out, or did he want to attempt them all personally? He gave up the engineering management role and concentrated on being CTO. The team got a dedicated leader. He got space to think.

“Are you ready to be fired?”

Grant Ingersoll, closing his 2016 essay on a founder’s dilemma

The joke has an edge. Technical competence can make a person useful in almost every room, which is a splendid way to spend an entire working day walking between rooms. Ingersoll’s career is full of contributions that travel further when other people can use them: open code, a practical book, training, interviews, and organizations built around shared knowledge.

Before the search box, FORTRAN

His first paid programming job involved computational electromagnetics simulations, distributed using FORTRAN, at a small company in Syracuse, New York. The work also took him into video conferencing and distributed Linux systems. That is a rather different opening scene from the familiar startup origin involving a dormitory and an impatient investor. There were simulations to run.

He studied mathematics and computer science at Amherst College, then earned a master’s in computer science at Syracuse University. At Amherst, his activities included varsity ice hockey. Later biographies would list cycling, climbing, hiking, and time with family among his interests. The person behind the search algorithms had reasons to close the laptop.

Search and language became a sustained occupation. His work at Syracuse’s Center for Natural Language Processing included cross-language information retrieval and an English question-answering system. These are practical encounters with the awkwardness of language: a useful document may be written differently from the query, or in another language altogether. Getting the words into a computer is only the beginning.

Open-source development gave that work a wider community. Ingersoll became a longtime contributor to Lucene and Solr and a co-creator of Apache Mahout. In a 2010 ApacheCon conversation, the subject was making machine learning available to a wider audience. Recommendation systems provided a recognizable example. The mathematics mattered; so did giving developers a way to put it to work.

The business around the code

Lucid Imagination, later Lucidworks, was formed in 2007. In August 2011, the company promoted Ingersoll to chief scientist, with responsibility for research direction and product architecture. His public explanation of the opportunity centered on organizations struggling to make use of their accumulating information. Search, language processing, and machine learning belonged in the same conversation.

The business came with an interesting condition. It built on Apache projects whose contributors did not all work for Lucidworks. A company could make plans; the community could make different ones. Contributions from elsewhere might even resemble features the company hoped to sell. Software development had acquired the diplomatic possibilities of a shared kitchen.

Ingersoll’s 2015 account of the business described selling a commercial product and enterprise support while contributing to the underlying projects. The company also sold expertise. Customers who had basic search running wanted help with harder problems, including language processing and feedback. Their questions moved beyond whether the engine worked toward what they could do with it.

That distinction helps explain his continued interest in search. A working installation is an intermediate result. Someone still needs to find the document, understand the customer, or decide what to do next. Technical success can leave a substantial amount of human work unfinished.

A book, a game, a better explanation

In December 2012, Manning published Taming Text: How to Find, Organize, and Manipulate It. Grant S. Ingersoll wrote it with Thomas S. Morton and Andrew L. Farris, with a foreword by Liz Liddy. Its 320 pages covered search, recognizing names, clustering, tagging, extracting information, and summarization through practical examples.

The title suggests a firm hand. The subject rewards patience. Text arrives with ambiguity built in, and the book’s scope acknowledges that searching it is part of a larger job of organization and interpretation. Readers were being introduced to a set of tools they could combine, with examples to make the abstractions less slippery.

His conference appearances found other ways into the subject. At Berlin Buzzwords in 2016, his talk was called “Level Up: Everything you need to know about search, you learned playing video games.” In 2017, “BM25 is so Yesterday” promised a discussion of relevance tuning, learning to rank, and query intent classification. The playful titles led into specific engineering problems.

The teaching continues through collaborations. With Daniel Tunkelang, he developed courses on search fundamentals and search with machine learning; with Dave Anderson, search engineering. The public engineering-course repository contains exercises and configuration across four weeks. A lesson becomes more useful when a student can run something, change it, and see what breaks.

The infrastructure of a shared answer

In September 2019, Ingersoll joined the Wikimedia Foundation as CTO. The foundation operates Wikipedia and its sibling free-knowledge projects. He would lead technical platform strategy across a global department whose work included engineering, research, security, machine learning, and analysis. The mission gave his familiar subject a different public setting.

On a February 2020 podcast, his role was described as leading a 120-person team, with continuing opportunities to learn from the people around him. The conversation covered Wikimedia’s knowledge infrastructure and Wikidata. A search specialist had entered an organization where knowledge was also the collective undertaking of volunteers, editors, researchers, and readers.

In another discussion about technical debt and scalable systems, he advocated experiments and incremental changes. It is a patient way to talk about an organization used by vast numbers of people. Infrastructure has to permit new work while carrying the existing work. Readers generally notice the arrangement when it stops working, an unglamorous but persuasive measure of its importance.

His Wikimedia tenure ended in July 2021. The foundation’s transition announcement thanked him for his service to free knowledge and the movement. The chapter fits his wider career without requiring every job to have been the same: commercial search and a public knowledge institution ask different things of a technology leader, even when both depend on people reaching information.

Careers refuse to travel in straight lines

Alongside the executive work, Ingersoll began collecting other people’s career stories. The Develomentor podcast launched in 2019 and recorded more than 130 conversations before winding down in 2022. Guests included founders, engineers, researchers, and people who had entered technology from music, bootcamps, and lumberjacking. A career ladder would have needed several improbable extensions.

He was learning a new craft himself. In a short account of recording early interviews, he thanked Drew Farris for being the first person to sit down with him. The audio quality needed improvement. The encouraging discovery was simpler: someone would sit across a microphone and talk about a career. For a new interviewer, that is quite enough to build on.

The show’s premise made room for choices that a résumé compresses into dates and titles. Its conversations covered roles across software engineering, data science, management, and sales engineering. The questions included compensation, freelancing, and joining startups. Listeners could hear how people made decisions rather than merely inspect where they ended up.

Back to the work, with more people in it

Develomentor was incorporated as a side project in 2018 and became a full-time firm in 2022. Today it supplies fractional technology, product, and data leadership, alongside search and AI work and technical due diligence. The podcast’s name stayed; the work acquired different obligations, including decisions that clients have to live with after the meeting ends.

His own public reflections keep returning to curiosity and feedback. In a LinkedIn post, he described preferring guided exploration and following an interest far enough to see whether it held. He also recalled a conversation with mentor Jon Williams about executives learning from feedback, including the kind that remains implicit unless someone makes the effort to surface it.

There is a pleasing continuity here. Search engineering asks what signals reveal about what people need. Interviewing asks someone to explain a decision. Leadership asks whether the people around you can tell you something useful, and whether you will listen. Ingersoll has spent a career moving among those questions. A good answer still requires somebody to pay attention.