Applied artificial intelligence · Consulting firm
Algognitive is a consulting firm dedicated to artificial intelligence. We work with companies that want to turn domain expertise into working systems, from initial scoping through to production.
We do not sell a platform and we do not resell someone else’s technology. We work on the client side, with a single objective: technology that produces a measurable result in your business.
2019
Firm founded
30
Years of applied AI, research to production
6
Sectors of recurring engagement
10+
Countries covered on deployment work
Founded in 2019, drawing on thirty years of research, development and real-world deployment of artificial intelligence.
Software · Media (publishing, music, broadcast, film) · Tourism · Real estate · Architecture
01
Use-case scoping, value and feasibility assessment, build versus buy, technology roadmap. Many AI projects fail before the first line of code, because the problem was framed wrong.
02
Architecture, data governance and quality, algorithm design, industrialization. We stay involved through production, because that is where the real gap between a prototype and a product shows up.
03
Innovation management, technology spin-offs, optimization of existing systems, international rollout. Technology that does not spread through the organization creates no value.
Our approach
AI is no longer constrained by technology availability. Models, tools and compute are accessible to everyone. The difficulty has moved elsewhere: choosing the right problems, getting data quality right, integrating into real processes, and earning the teams’ buy-in.
01
An engagement begins by understanding how the work is actually done, what it costs, where it breaks, and what the best practitioners do without being able to explain it. Technology comes second.
02
A prototype proves something is possible. Production proves it is useful, robust, maintainable and accepted. We scope against real operating conditions: latency, inference cost, input data quality, error recovery, load, compliance.
03
Projects succeed when domain experts see the tool as a multiplier of their own value rather than a threat. That is built in from the design stage: formalizing know-how, involving the teams, drawing a clear line around what stays a human decision.
04
We would rather name a limitation than promise a result that will not hold. On AI work, an advisor’s credibility rests on ruling out the wrong paths as fast as possible.
Practice areas
A data transformation prepares an AI project, which produces a reusable component, which becomes a spin-off, which then scales internationally.
Before the algorithm comes the data. We work across the full chain: source mapping, quality, data model, collection and storage architecture, governance, GDPR and AI Act compliance.
Design and delivery of AI-based systems: machine learning, combinatorial optimization, language models, recommender systems, vision, signal processing. We work with statistical, symbolic and hybrid methods, chosen on the merits of the problem rather than on fashion.
Choosing an architecture commits years of cost and capability. We help executive teams and investment committees decide: technical foundations, dependency on model providers, data sovereignty and location, total cost of ownership, regulatory exposure.
Many companies hold a technology asset with value beyond their own use without realizing it: an optimization engine, a reference dataset, a formalized method. We identify those assets and build the path that turns them into a product, then into a standalone company.
Building a repeatable capacity to innovate rather than running experiments that go nowhere. Project portfolio, decision criteria, the handoff from R&D to product, the relationship between research and operating teams, measurement of value created.
The historical core of our practice: expressing an operational problem as a mathematical one and solving it. Planning, resource allocation, routing, scheduling, pricing, space allocation, inventory. These projects produce direct, measurable gains, often faster than generative AI projects.
Going from one market to ten is not a translation exercise. Product localization, multi-country architecture, compliance, distributed team structure, commercial model adaptation, entry market selection.
Selected work
We work with technology companies over the long term, sometimes as advisors, sometimes as co-founders. In several cases the engagement extended into a co-founder or board role.
AI decision-support for workplace design
Waitack generates complete, realistic and compliant layout plans in seconds, where the traditional method takes days to produce a single proposal. Users compare scenarios in real time across four dimensions: human, environmental, operational and financial.
Publisher of Castopod, the open-source podcast platform
Castopod is a podcast hosting and distribution platform released under a free license and built on the Podcasting 2.0 and ActivityPub open standards. It is used by television channels, FM radio stations, universities and press groups across roughly thirty languages.
Energy decarbonization through electric mobility
illigo works on the deployment of electric mobility and the financing structures that make it economically viable.
About
Practice of AI since 1994 · industrial optimization · logistics · transport · recommender engines · semantic search
Algognitive was founded in 2019 on a simple observation: a gap had opened between how mature AI technologies actually were and how well companies could use them. Not for lack of tools, but for lack of translation between domain expertise and technical architecture. Our work is exactly that translation, in both directions — understanding a business well enough to formalize what its best practitioners do without thinking about it, and knowing the technology well enough to say what it can carry, at what cost, and with what limits.
Our practice starts in 1994, well before AI’s current diffusion. That continuity has a practical consequence: we have watched several technology cycles come and go, and we can tell a genuine break from a renamed version of an existing method.
Small teams, drawing on a network of specialists brought in by subject: mathematicians, data engineers, software architects, domain experts. We would rather have three people who understand the problem than fifteen executing a specification.
On intellectual property the principle is straightforward: developments, algorithms and creations produced during an engagement are assigned to the client, contractually, from the start.
The founder
Engineer, graduate of École Centrale Paris and AgroParisTech, specialized in applied mathematics and artificial intelligence since 1994.
CIO and president of subsidiaries at BNP Paribas Immobilier · president of Périclès, a real estate software vendor whose revenue grew from €1M to €5M before its sale to seloger.com · founder and CEO of the Keeward Group, present in ten countries with more than 250 employees and €60M in revenue across software, media and online commerce, used as a Harvard Business School teaching case · executive director for R&D, technology and marketing at a workplace design and build firm · co-founder and CEO of Waitack · author, including the novel Le parti de l’homme (Tamyras, 2014).
Contact
The first conversation is about qualifying the problem, not selling an engagement. Describe your situation in a few lines and we will tell you quickly whether we are the right people, and if not, who is.
Do you take on projects already under way?
Yes, often. An outside view is usually more useful on a project in trouble than on a blank page.
Do you work with SMEs or only large groups?
Both, as well as start-ups moving into technical structuring. Size matters less than clarity of the stake.
Do you supply development teams?
We do not do staff augmentation. We design, lead, and help you hire or select delivery partners.
How long is a typical engagement?
From a few weeks for a scoping exercise to several years for a fractional technical leadership or board role.
Reply within two business days