Evidence, then decisions.
Trajanov is an applied-AI consultancy for organisations that answer to a board. We take one slow, manual process at a time (reports, documents, decisions) and build the system that does it in hours: on your infrastructure, owned by your team, priced in plain numbers.
- 200+
- Peer-reviewed papers
- 70+
- Projects delivered
- 20+
- Years at the frontier
- 3
- Continents in practice
Figures as counted in 2026: publications, delivered engagements, years since first academic appointment, continents where the firm practises.
The position
Most AI programmes stall between the slide deck and the working system. Ours are designed, and priced, not to.
Trajanov is three people who share a name and split a problem: a professor who has shaped AI research and policy for two decades, and the engineers who turn that research into systems that run in production. Strategy and shipping under one roof, and one surname on both.
A working example: a report your team assembles by hand over three days comes out of the system in two hours, on your own data, in your own brand. That is the shape of everything we build — one slow manual process, made automatic, measured honestly. Begin small, scale what earns it.
“We would rather tell you AI is the wrong tool than sell you a model you don’t need.”
What we build
Three services, prices visible. The whole ladder in one line: €1,000 to prove it works. €6,000 to build it. €2,000 a month to keep it working and keep improving it.
-
S.1
The Pilot
€1,000 · two weeks
Give us one report your team makes by hand. In two weeks you see the same report made automatically, with your own numbers in it. The fee comes off the Build if you continue.
-
S.2
The Build
€6,000–9,000 · four to six weeks
The finished system, running on your systems, not a demo. Your team trained, everything documented, two rounds of changes included. It keeps working after we leave.
-
S.3
The Retainer
€2,000–3,500 / month · three-month minimum
We keep it running, fix what breaks, and every month we automate one more thing that eats your team's time. Systems need someone; that is the honest reason retainers exist.
Under every system: language models and generative AI (including 4B-parameter models we trained ourselves), agents and automation, data science, NLP and knowledge graphs, and evaluation that stands up to a regulator, and to your board.
Statement of standards
Modelled on academic rigour and production discipline. Every engagement leaves you with the code, the documentation, the evaluation suite, and a team trained to own them.
Reproducible results
If we can’t run it twice, we don’t report it once.
Documented decisions
Every trade-off written down, so the next team inherits reasons, not folklore.
Measurable outcomes
Agreed metrics before we build — and honest numbers after.
A hard line on claims
We state what AI can do for you, and we put in writing what it cannot.
Where we practise
An index of the sectors in which the method has been applied. Any industry where data meets decisions qualifies.
- Healthcare & Pharmasee: diagnostics, drug discovery, clinical NLP
- Finance & FinTechsee: risk, fraud, robo-advisory, agents
- Energysee: forecasting, optimisation, monitoring
- Public Administrationsee: digital services, document AI
- Agriculturesee: remote sensing, yield models
- Language & Culturesee: low-resource LLMs, translation
- Retail & E-commercesee: recommendation, demand, search
- Climate & Sustainabilitysee: modelling, ESG analytics
The Trajanovs
Dimitar Trajanov, Ph.D.
Plate IPrincipal AI Advisor
Full professor at FCSE Skopje and visiting research professor at Boston University. Founding dean, associate member of the Macedonian Academy of Sciences & Arts, and leader of Vezilka — North Macedonia’s national AI Factory. 200+ papers, 70+ projects, 7 books.
Risto Trajanov
Plate IIData Scientist · ML Engineer
M.Sc. in Data Science, Rice University, as a Fulbright Scholar; data scientist at Deutser. Builds machine-learning systems that ship: report automation that returns dozens of hours to teams, ETL pipelines, PyTorch models in production. Six peer-reviewed papers.
Darko Trajanov
Plate IIIFull-Stack & AI Engineer
The builder who takes models out of the notebook and into the browser: the interfaces, APIs and infrastructure that turn prototypes into products your people actually use.
Notes & questions
Note 1Do we need our own data to start?
Helpful, but not required. We can begin with a discovery sprint, work with sample or public data for a prototype, and design the data collection you’ll need as part of the engagement.
Note 2How fast can we see something real?
Usually a working proof of concept on your real problem within a few weeks. We deliberately front-load the riskiest assumptions, so you can decide to continue, or stop, with evidence rather than a pitch.
Note 3Will you tell us if AI is the wrong tool?
Always. A good part of our value is telling you where a simpler solution wins, or where the data isn’t ready yet. We’d rather keep your trust than sell you a model you don’t need.
Note 4Do you hand over, or hold the keys?
We hand over. You get the code, the documentation, the evaluation suite and the training your team needs to own and extend the system confidently.
Note 5Can’t we just use ChatGPT?
For drafts, yes — and you should. A system is different: it runs on your data, in your format, every time, without anyone typing. Where a chat window is enough, we will say so and step aside.
Note 6What happens when a model is retired?
Models get retired, data formats change, new reports get requested: a system needs someone. That is what the retainer is. If you would rather own that yourself, the handover includes everything your team needs.
