Ulaanbaatar 47.92°N 106.92°EDentsu GroupAlways to create high value

AI alone isn't enough.

Adaptive Intelligence

DDAM is the AI and data engineering arm of Dentsu Digital. We design the models, pipelines and systems — and our clients run their businesses on them.

Everyone can reach the same models now. The difference is who can get them running inside a business: tested, deployed, and still working next quarter.

Orientation Know where you are Time Know when you are Movement Know when to move
Start a project What is an örtöö?

ÖrtööDDAM is the relay station at the centre. Nine things we actually do orbit it — pick one to see where it leads.

Explore the network — choose a station0 / 9 reached

ST 01The idea

ӨРТӨӨ

Örtöö — the relay station. Mongol Empire, thirteenth century. Proof that implementation is the hard part.

  • Know where you are.Orientation
  • Know when you are.Time
  • Know when to move.Movement

Three things a rider had to have before the message could leave. They are also where DDAM’s own design principles start — four more carry the same logic further.

  • Khana
  • Tension
  • Portability
  • Nomadic Intelligence

Mongolia built the fastest information network in the world and held the record for centuries. Relay stations roughly a day apart, from the Danube to the Yellow Sea: a rider came in exhausted, handed the message on, and a fresh horse carried it out.

The idea was not the achievement. Every empire wanted fast messages; none of them were short of the concept. What the Mongols actually did was build the implementation — where stations went, how horses were rationed, who counted as authorised, what happened when a rider failed.

The idea was ordinary. The implementation was the achievement. That is the same argument we make about AI, and it was proven here once already. We design the models and the systems; our clients run their businesses on them.

We don't deliver the work. We build the thing that does it.

You're carrying the light

Ask Damujin · Physical AI Interface

I carry messages across the relay for a living. Ask me anything about Örtöö or DDAM.

Who

Damujin is DDAM's character.

What he wears

A black deel, a gold sash, turquoise in his headband — and sneakers. The tradition is not a costume; it is what he wears to work.

Örtöö — the station

A day’s ride apart, all the way to the Danube.

No rider knew the whole route — only the next station.

Örtöö — the handover

Every station handed the message on, and let it go.

Nobody carried it the whole way. Nobody had to.

Örtöö — the pattern

That is the shape we build into now.

A system that keeps moving after any one rider steps off.

Relay — the carrier changed four times. The message never did.

Between 01 and 02

Raw in. Four things that did not exist before.

MovementNothing is carried unchanged. What arrives as noise leaves as something that runs.

Everything arrives as noise — logs, spend, impressions, half-labelled spreadsheets. We don't sort it and pass it on. We work out what it could become, then build that: a trained model, a pipeline, a proof, an automated operation. Our clients keep running them long after we hand over the keys.

01ModelsTrained, evaluated, deployed 02PipelinesConnected and governed 03ProofsSix weeks to an answer 04CampaignsRun, reported, automated
ST 02Capabilities

Four things we do, and we do them all the way to production.

No slide-ware. Every engagement ends with something running.

C/01

AI Solution Development

Models that survive contact with production.

Custom modules built for one business problem, not a demo — trained, evaluated, deployed and monitored. Recommendation, forecasting, classification, LLM tooling, computer vision.

Problem
Models that work in a notebook and fail on contact with production.
What we do
Train, evaluate, deploy and monitor a model against one business problem.
You receive
A running model, its evaluation record, and the monitoring around it.
Already done
Recommendation, forecasting, classification, LLM tooling and computer vision, in production.
MLOpsLLMForecastingCV
C/02

Data Engineering & Analytics

Data is strongest when it's connected.

Collection, cleansing, structuring, warehousing and the analysis on top. We build the pipeline that makes every later question cheap to answer.

Problem
Every new question costs a new project, because nothing is connected.
What we do
Collect, cleanse, structure and warehouse the data, then build the analysis on top.
You receive
A governed pipeline and warehouse your own team can query without us.
Already done
Cloud-native pipelines with governance, feeding the models above.
PipelinesCloud nativeBIGovernance
C/03

Proof of Concept & R&D

Find out in six weeks, not six quarters.

Most companies cannot say whether AI will help them until they test it. We run the hypothesis cheaply and fast, and we tell you plainly when the answer is no.

Problem
Nobody can say whether AI will help here until someone tests it.
What we do
Run one hypothesis against your own data, cheaply and in the open.
You receive
A plain answer in six weeks, and the number that decided it.
Already done
Six-week fixed-scope proofs — including the ones where the answer was no.
Rapid prototypingNo lock-inBenchmarking
C/04

Digital Marketing Operations

80% of campaign reporting, automated.

Front-end simulation for pitches, live campaign management, submission and creative control, and advertiser reporting for the Japanese market — with our own AI doing the repetitive part.

Problem
Reporting and submission work consumes the team that should be optimising.
What we do
Run campaigns, submissions, creative control and advertiser reporting.
You receive
An operated campaign, and the reporting, with the repetitive part automated.
Already done
80% of campaign reporting automated for the Japanese market.
Campaign opsJP marketAutomation

How the work actually runs

Collective intelligence, to progressively reduce uncertainty.

  1. 01Explore

    Bring together people, AI, data and market signals.

  2. 02Frame

    Define the question, hypothesis and what we need to learn.

  3. 03Simulate

    Explore possible reactions, scenarios and outcomes before acting.

  4. 04Build

    Create only what is needed to test the hypothesis.

  5. 05Test

    Put it in front of real people, workflows or environments.

  6. 06Learn & decide

    Use evidence to scale, change direction, repeat — or stop.

Sometimes the right outcome is a decision not to build.

ST 03Telemetry

The station, in numbers.

Headcount · 2018 → 2026 150 in 2026
2018 — FOUNDED2023 — INTO DENTSU DIGITAL2026 — TODAY
0 Specialists on staffOne floor in Ulaanbaatar, three languages 01
0× Headcount, past three yearsFifty people to a hundred and fifty 02
0% English proficiencyThe working language of the Global Division 03
0% Japanese proficiencyBriefs are read in the language they arrive in 04
0% Campaign reporting automatedOur own AI does the repetitive part 05
0% Senior & leadThirty-two in every hundred have shipped this before 06
ST 02bWorking with us

Turn an unfinished question into something testable — fast.

The traditional way

  1. Discussion
  2. Requirements
  3. Build
  4. Discover

The ORTOO Relay · DDAM Way

  1. Opportunity
  2. Scope
  3. PoC (6 wks)
  4. Build
  5. Production
  6. Operation

Six stations from first brief to autonomous client operations. The relay resolves uncertainty early, then delivers code you run.

Three shapes an engagement takes.

Every one is scoped and priced in writing before it starts. No open-ended retainer, no discovery phase that bills while nobody decides anything.

01 — Prove it

Proof of concept

One hypothesis, tested against your own data, with the number that decides it agreed up front.

Duration
Six weeks
Scope
Fixed
Fee
Fixed, quoted before we begin
You keep
Everything built, either way
02 — Build it

Delivery

A model, a pipeline or a platform taken to production and handed over documented, reproducible and yours to run.

Duration
Three to six months, typically
Scope
Milestoned, re-quoted if it changes
Fee
Per milestone
You keep
The system, the code and the documentation
03 — Run it

Operations

A standing team inside your programme — campaign operations, reporting, or a data platform kept alive and improving.

Duration
Ongoing, reviewed quarterly
Scope
Named team, agreed service levels
Fee
Monthly
You keep
Full visibility of what the team is doing
Case — our own operation

The reporting operation we automated was ours.

  1. Situation

    2023: we stood up digital marketing operations for Japan, starting with advertiser reporting — high volume, low judgement.

  2. Problem

    It scaled the way this work always does: linearly, with headcount. Every new advertiser meant another pair of hands.

  3. What we did

    We pointed our own AI and engineering at it: decomposed the production, automated the repeatable parts, kept people on what still needs one.

  4. Result

    80% of campaign reporting now runs without a person in the loop. It became its own division in 2024.

  5. What it means for you

    We sell this because we run it — the same engineering on a client’s operation as on our own floor. Not a deck. Load-bearing.

80%of campaign reporting automated
2023operations launched
2024became a division
Ask how it applies to you
ST 02cBuilt here
Product

People Model

People ModelIn production

A hundred million personas modelling Japan’s consumer population — grounded in real survey data, and able to answer a research question in minutes rather than weeks.

Ask a large model to imagine a Japanese eighteen-year-old and it will answer fluently and be wrong in ways you cannot detect. That is the whole problem in one sentence, and it is why the interesting engineering is not the generation.

  • Survey-grounded skeleton
  • Demographic filters
  • Natural-language query
  • Segment comparison
  • Evidence per answer
How it works Ask about it
One persona — drawn, not imagined
100M personas — the scale of Japan’s consumer population

Grounded in repeated national survey waves, not generated from a prompt. Filter to a prefecture, an age band or a segment and query that population directly.

Two thousand of a hundred million

Every view below is one population, not five charts.

Six stations on one route. At each one the same two thousand people are re-sorted, never replaced — the arrangement changes, the population does not.

Shape from published national marginals. Individuals illustrative.

Built to explore what comes next

Our platforms and labs turn questions into experiments, and experiments into real impact.

Agents for a bigger tomorrow DDAM Canvas

Build and use purpose-specific AI agents.

Ideas, people, possibilities AI Studio

A space for collective intelligence.

Explore before you act Simulation & Synthetic Intelligence

Explore people, scenarios and possible outcomes before acting.

Intelligence in a wider world Physical AI Lab

Explore what happens when AI enters the physical world.

Local insights. Global impact.

Mongolia — real-world proving ground

Connect experiments with real consumers, brands, environments and market conditions.

Real people
Real places
Real progress
Between 03 and 04

A hub is only useful if it reaches somewhere.

RouteLegs between stations, and a hand-over at each one. The message never stops.

The regions our president names as the network Mongolia connects: the UK, Singapore, India, Taiwan, China, Indonesia and Vietnam, with Japan as the group’s home. Plotted from Ulaanbaatar with real great-circle distances and time offsets — Tokyo is 3,008 km and one hour ahead, near enough that a question and its answer happen the same morning. Spin the globe. Click a station.

APAC

Tokyo

Distance3,008 km Offset+1 h
Trajectory of Value & Knowledge

Global does not mean headcount. It is problems, knowledge and technology moving between time zones and arriving as real value. Click a corridor to focus the globe.

Station 01 · Hub
UlaanbaatarThe Engine · 150 Specialists
Where it starts
Altan Joloo Tower 6F — every global relay is received, engineered, and monitored here.
What DDAM builds
Custom ML models, 100M persona groundings, governed pipelines, and 6-week proofs.
Value delivered
Zero slide-ware: running models and production systems shipped across 3 continents.
Tokyo → UB → Tokyo
Tokyo — AI Solutions3,008 km · +1 h
Market challenge in
Japanese enterprise challenges, client briefs, and massive telemetry ad logs.
What moved & engineered
Algorithm squads train, evaluate, and calibrate custom ML & recommendation models.
Production result
80% of campaign reporting automated; deployed directly into Dentsu client environments.
Bengaluru ↔ UB
Bengaluru — Data Scale4,738 km · −2.5 h
Infrastructure in
Distributed cloud-native data warehousing architectures across the Dentsu network.
What moved & engineered
High-throughput data pipelines, automated schema governance, and 24/7 MLOps telemetry.
Production result
Autonomous data warehouses that client engineering teams query without ongoing lock-in.
London ↔ UB
London — Frontier AI6,972 km · −8 h
Innovation brief in
Frontier agent systems, spatial computing, and international creative tech concepts.
What moved & engineered
6-week fixed-scope proofs of concept and Physical AI prototypes tested on Mongolian ground.
Production result
Unambiguous numerical proofs with validated go/no-go decisions before client capital spend.
ST 04The network

A Mongolian company with a Tokyo mandate.

Data Artist was integrated into Dentsu Digital in April 2023. DDAM became a subsidiary of Dentsu Digital Inc. — which makes us a working part of the Dentsu Group, not a vendor at the edge of it.

In practice our engineers help shape the programme rather than receive it: we bring the R&D, we propose what should be built, and what we build gets adopted across the group and ships under a name that has been accountable in that market for a hundred and twenty-five years.

Talk to the Global Division
GroupDentsu Group Inc.Tokyo · founded 1901
ParentDentsu Digital Inc.Japan's digital marketing practice
Merged 2023Data Artist Inc.AI & data science
You are hereDentsu Data Artist Mongol LLCUlaanbaatar · AI, data engineering, marketing ops
Registered name
Dentsu Data Artist Mongol LLC
Founded
June 2018
President
Hatsumi Imai, President and Executive Officer
Headquarters
Altan Joloo Tower 6F, Seoul Street, 5th khoroolol, 3rd khoroo, Sukhbaatar district, Ulaanbaatar 14252, Mongolia
Parent
Dentsu Digital Inc., Tokyo
Group
Dentsu Group Inc., Tokyo — founded 1901
Headcount
150 specialists

From possibility to evidence and into reality

  1. 01Simulate before you act

    Explore possible reactions, scenarios and outcomes before committing real resources.

  2. 02Build while the question still matters

    Turn an unfinished idea into something people can see, use and react to — fast enough to change the next decision.

  3. 03Bring AI into the real world

    Explore what happens when intelligence moves beyond a conventional screen.

  4. 04Use the market as proving ground

    When the question requires real people, real environments and real behaviour, Mongolia can become a place to learn.

  5. 05Transform the work itself

    We do not stop at prototypes. We use AI inside real operations, and learn from what happens next.

ST 04bPresident’s message
Hatsumi Imai, President and Executive Officer of Dentsu Data Artist Mongol
Hatsumi Imai President and Executive Officer

An AI development hub for the dentsu group, built in Mongolia.

We are living in an era where generative AI is rapidly transforming the way businesses operate, create value, and compete on a global scale. As technology continues to evolve at an unprecedented speed, companies are expected not only to adopt new tools, but also to rethink how they work, collaborate, and deliver meaningful impact.

At Dentsu Data Artist Mongolia (DDAM), our ambition is to become a global AI development hub for the dentsu group. Based in Mongolia, DDAM brings together strong capabilities in digital advertising, BPO, data, engineering, and AI development. By combining operational excellence with advanced technology, we support business transformation and contribute to the growth of Dentsu Digital, dentsu Japan, and the broader dentsu group.

Hatsumi Imai · President and Executive Officer

Management team

Makito Tsukahara, Vice President and Executive Officer
Makito TsukaharaVice President and Executive Officer
Yoshiki Miyamoto, Executive Officer, Corporate Planning and Administration Division
Yoshiki MiyamotoCorporate Planning and Administration Division, Executive Officer
Khandmaa Batbayar, Executive Officer, Digital Marketing Division
Khandmaa BatbayarDigital Marketing Division, Executive Officer

What DDAM did

Result or learning

ST 05The route so far

Eight years, one direction.

Keep scrolling — the route runs sideways

2015

Data Artist, Tokyo

The parent company starts building AI for marketing in Japan.

2018.06

DDAM is founded

An R&D and development centre opens in Ulaanbaatar, and joins the Dentsu Group.

2019

Beyond marketing

AI work expands into HR, education and other sectors.

2023.04

Into Dentsu Digital

Data Artist is integrated into Dentsu Digital Inc.; DDAM becomes its subsidiary.

2024.01

Digital Marketing Division

Campaign operations stand up as a division of their own.

2025.09

Global Division

A dedicated route for clients outside Japan.

2026.04

150 people, new structure

Tripled in three years. Management moves to an executive officer system.

Next

The station after this one

Deeper AI in operations, more work run end-to-end from Ulaanbaatar.

ST 06The building

Six rooms that argue culture and technology aren't opposites.

Our floor in Altan Joloo Tower was designed around four principles — technological expression, deep focus, multicultural integration, and architecture that assumes the future arrives. It is, we think, the most comfortable office in Ulaanbaatar. Come and disagree.

A full Mongolian ger built inside the DDAM office, surrounded by a painted mural of Mongolian history
Mongolian Area

The ger

A real ger on the floor, ringed by a hand-painted history mural. Where visiting clients get tea before anyone opens a laptop.

Event area with blue neon ceiling lighting, DD lettering and pop-art panels reading Always To Create High Value
Japan Area

The Great Wave, in circuitry

Japanese motif redrawn as a printed-circuit pattern, under cyberpunk neon. Doubles as the event stage.

Open-plan DDAM workspace with long white desks, monitors and circular acoustic ceiling panels
Work Space

The open floor

Open zones for daily work, teamwork, heads-down focus and international meetings — four modes, each with somewhere to go.

Dark boardroom with linear lighting and an illuminated blue humanoid AI sculpture on the wall
AI Studio

Technology & AI Lab

Compute, smart presentation suites, and the room where proofs of concept get their verdict.

Knowledge hub with arched library alcoves, ornamental panels and a red British telephone box
APAC Area

The Knowledge Hub

Library alcoves for deep work and continuous learning. Yes, that is a telephone box.

Dark British bar-style social lounge with marble counter, cane bar stools and warm pendant lighting
Social

The DDAM Pub

British bar-style lounge for informal syncs, the end of a hard sprint, and the conversations that don't fit in a meeting room.

Between 06 and 07

“People are the nodes that make ORTOO work.”

NodeA station is a place with someone in it. Empty, it is only a building.

Here is our office floor in Altan Joloo Tower, photographed with nobody in it — desks, monitors, and chairs. That is the hardware any company can assemble. Move across it: the space only activates where you look, and a light illuminates at every seat you reach. A station is only a shell until someone sits at the desk. Every seat is a node: people are the nodes that make the ORTOO relay work.

32%Senior & leadHave shipped this before 68%Junior & growingBeing taught by them, on live work 3×Growth in three yearsAnd still hiring
ST 07People

Grow together. Create value. Belong as one team.

We are a company that runs on human intelligence, so the people come first — mentored by specialists who have worked with global clients, practising the soft skills as deliberately as the hard ones.

What they know
  • Machine learning engineering and MLOps
  • Data engineering, warehousing and governance
  • Analytics and business intelligence
  • LLM tooling and evaluation
  • Computer vision
  • Campaign operations for the Japanese market
What they build
  • Trained models, deployed and monitored
  • Governed pipelines and warehouses
  • Six-week proofs of concept
  • Automated campaign reporting and submission
  • Our own products — People Model among them
What they are exploring
  • Agent systems that carry out multi-step work
  • Physical AI — models that act in real environments
  • Simulation before commitment
  • Mongolia as a live proving ground
01 — Learn & grow

Always learning.

Continuous learning, knowledge shared out loud, and new ideas given somewhere to land. Training, certifications, and a path to the next stage.

02 — One team

We achieve more together.

Different perspectives, actually listened to. Collaboration with people from backgrounds and disciplines unlike your own.

03 — Belong & connect

Everyone has a voice.

An open, inclusive floor where people can be themselves, stay connected, and shape the place they work in.

Build the future of technology

Real AI and digital-marketing problems from real businesses, not sandbox exercises.

Two markets, one career

Work across the Japanese and Mongolian technology landscapes and pick up genuine global experience.

Certified, not just busy

Structured training, professional certification, and L&D budget that gets used.

Do great work, live well

A modern floor, wellbeing programmes, and a schedule that assumes you have a life.

Junior AI / ML / Data EngineerDesign, train and test ML models for digital marketingFull time · UB→ Software EngineerReliable, secure systems with automated testing and real design practiceFull time · UB→ Digital Marketing Report SpecialistDaily and weekly campaign reporting for the Japanese marketFull time · UB→ Digital Marketing Report SpecialistSame work, flexible hoursPart time · UB→
ST 07cBefore you write to us

Questions people ask first.

You are in Ulaanbaatar. Does the distance matter?

Tokyo is 3,008 km away and one hour ahead. A question asked in the morning is answered the same morning, which is the part that actually matters — not the kilometres. Offshore arrangements fail on time zones, not geography.

Can you work in Japanese?

Seventy per cent of the floor works in Japanese and ninety per cent in English, alongside Mongolian. Briefs are read in the language they arrive in, and our engineers sit inside Japanese client programmes rather than receiving translated summaries.

What does a first engagement look like?

Almost always a proof of concept: six weeks, one hypothesis, fixed scope and a fixed fee agreed before it starts. If the honest answer at the end is that it is not worth building yet, you get that answer — and everything built along the way.

Who owns what you build?

You do. Code, models, pipelines and documentation are handed over so your team can run and change them without us. We do not build dependencies on ourselves.

Are you an agency, or part of Dentsu?

Dentsu Data Artist Mongol LLC is a subsidiary of Dentsu Digital Inc., which is part of Dentsu Group Inc. Work ships under the group’s governance, security and accountability standards.

How quickly will someone reply?

Write to ddam@group.data-artist.com and a person — not a form robot — reads it. The first reply tells you who owns the conversation and what the next step is.

ST 08Build it

Bring us the unfinished question.

  • A problem
  • An idea
  • A hypothesis
  • A strange possibility

Tell us the problem, not the spec. If a proof of concept is the honest first step we will say so, and we will have run it before the quarter is out.

Start a project See open roles
Emailddam@group.data-artist.com
Phone+976 7711 3326
Office

Altan Joloo Tower 6F, Seoul Street
5th khoroolol, 3rd khoroo, Sukhbaatar district
Ulaanbaatar 14252, Mongolia