Data-intensive systems · founder-led

Keep what works.Fix what doesn't.Build what's missing.

Solve complex operational problems without turning them into complex IT projects.

Explore products

years in software: 40+; years in software
uptime in production: 99.998%; uptime in production
direct engineering: 1:1; direct engineering

datrmine — execution plan

PROJECT / OPERATIONAL CHANGE6 steps

01existing systemsKEEP
02missing capabilityBUILD
03broken workflowFIX
04edge casesHANDLE
05unnecessary overheadSKIP
06productionLIVE

STATUSREADY FOR PRODUCTION

Before building

What looks like a project

Looks like: A database projectOften turns out to be: One missing data flowHistorical crew data ↓
Looks like: A new enterprise platformOften turns out to be: A bridge between existing systemsTowage crewing ↓
Looks like: Months of specificationOften turns out to be: One missing workflowVoice agent ↓
Looks like: A replacement projectOften turns out to be: Something worth leaving aloneExisting phone system ↓

„The project is often bigger than the problem.“

Products

Built for operations, proven in production

Software shaped by real operational work — starting with towage crewing.

Towage Crewing & Competency

Not just where a tug is, but whether it can go.

For towage operators: crew, qualifications and certificates at a glance, so you know instantly whether a tug is ready for the job. Built around the departments' existing systems.

Platform details (coming soon)

In production Europe-wide at FAIRPLAY TOWAGE GROUP

Readiness on the map
Every tug in context, directly on the harbour map: positions, crews on board and open warnings in one view.
Competency in one click
From the vessel straight to the crew member: certificates, validity dates, scanned originals, experience, history and availability.
From plan to deck, in sync
Planners assign, masters confirm, crew sees their next rotation — all on the same live data.

fleet · port of hamburglive

loading map…

ELBE-07 · Hamburg - Germany

30.07.26 – 18.09.26

Crew · rank

CW31

CW32

CW33

CW34

CW35

CW36

A. MercerMaster

J. ValeMaster

S. MarinChief Engineer

L. NovakChief Engineer

T. ReedDeckhand

K. BrandtShip's Mechanic

OnboardTime offVacationTrainingSick leave

6 days onboard

Member: J. Vale (Master)
Period: 13.08 – 18.08
Vessel: ELBE-07
Exact: 13.08.26 06:00 – 18.08.26 18:00
Reminder: not sent
Remarks: Relief required — master reported sick.

Cancel

Reminder

Remove

Save

Click an assignment in the plan · all data anonymised

AI Voice Agent

Keep your phone system. Automate the work behind the call.

No new telephony, no new CRM. The agent joins your existing PBX, works through your own APIs and hands anything that needs a human to your ticket system.

3 meetings from first conversation to production.Existing telephony stayed in place.

In production at PCA

Joins your phone system
Registers as an IP phone on your PBX. All SIP-based systems work. The dialled extension decides the tenant.
Workflows are configuration, not code
Services, rules and API calls defined in JSON, with linting and built-in help. Any company with an API can plug in.
Done, or queued for a human
Tasks are executed on the call or land in your ticket system. Once staff confirm, a webhook lets the agent finish the job.

call · ext 210live

Voice agentOnlineOpenAI Realtime · voice Cedar
TelephonyRegisteredIP phone on the existing PBX
Customer APIConfiguredper tenant, freely mapped
TicketsConnectedreview and confirm via webhook

parallel calls · 0 / 203 tenants · routed by extension

0 / 11 steps

schema valid

1{2  "Goal": "Identify the affected child as the customer and provide the customer number for subsequent services. Parents are callers, not customers.",3  "Prerequisites": [],4  "Communication": {5    "Instructions": "Greet the caller first. If knownPhone=true, ask for the child's first name; otherwise ask for contract number and date of birth.",6    "Rules": [7      "Only use details the caller has given or that are already in context.",8      "Never invent customer data.",9      "On multiple matches, never pick one yourself.",10      "Run no other service before the customer is identified."11    ]12  },13  "Inputs": [14    {15      "Name": "contractId",16      "Type": "integer",17      "Required": false,18      "Description": "Contract number of the child."19    },20    {21      "Name": "birthDate",22      "Type": "date",23      "Format": "yyyy-MM-dd",24      "Required": false,25      "Description": "Date of birth of the child, not the parent."26    },27    {28      "Name": "firstName",29      "Type": "string",30      "Required": false,31      "Description": "First name of the affected child."32    }33  ],34  "Api": {35    "Config": "API",36    "Endpoints": [37      {38        "Id": "customer_lookup",39        "Method": "POST",40        "Path": "/customers/lookup",41        "Parameters": [42          {43            "Name": "contractId",44            "Input": "contractId"45          },46          {47            "Name": "birthDate",48            "Input": "birthDate"49          },50          {51            "Name": "firstName",52            "Input": "firstName"53          }54        ]55      }56    ]57  },58  "Evaluation": [59    {60      "Result": "UniqueMatch",61      "Condition": "Exactly one customer is returned."62    },63    {64      "Result": "MultipleMatches",65      "Condition": "Several possible customers are returned."66    },67    {68      "Result": "NoMatch",69      "Condition": "No customer is returned."70    },71    {72      "Result": "TechnicalError",73      "Condition": "The API call fails."74    }75  ],76  "ContextUpdates": [77    {78      "Name": "CustomerNumber",79      "Source": "customer_lookup.Response.customerId",80      "Required": true81    }82  ]83}

Goal is set
Every input has a name and type
Endpoint ids are unique
Parameters map to known inputs
TechnicalError is handled
Context sources reference endpoints

Pick a call or a service · all data anonymised

AI where it pays off, inside your own systems

Not every problem fits a product. Custom AI and automation, integrated with your existing schemas, designed for reliability and built to be maintained long term.

Experience

Built to last

FAST has been in production since 1995 for broadcasters, publishers and media marketers: more than 1.2 billion advertising placements across every major medium, complex ad-hoc reports in about 1.4 seconds, and Tableau dashboards embedded right in the application. That track record is why we can build serious operational software today.

In continuous production: Since 1995
Advertising placements captured: 1.2 billion
Typical complex ad-hoc report: ~1.4 s

Frame 1 of the spot — Brand and figures shown are illustrative.
Print
TV
Online
Streaming
Cinema
Out-of-home
Direct mail

Frame 2 of the spot — Brand and figures shown are illustrative.
Print
TV
Online
Streaming
Cinema
Out-of-home
Direct mail

Spend per TV channel for this spot
Last	Channel	k€
	Total	4,180
28.06.26	RTL	1,240
28.06.26	ProSieben	1,010
27.06.26	Sat.1	780
28.06.26	VOX	520
26.06.26	Kabel Eins	340
27.06.26	n-tv	290

Selected organizations over the years

ARD
RTL
Sat.1
Axel Springer
Burda
G+J
DER SPIEGEL
S.Z.
Handelsblatt
IQM
Visoon
Bauer Verlag
MPS
B&M
Lotto24
WPV
Roche
pilot
Initiative Media
rs media
WEFRA
GFI
Biermann
Avoxa
Financial Times
mds
Medical Tribune
Urban+Vogel
FRoSTA

Integrates with

Tableau, embedded
Excel
PowerPoint
Qlik
Alteryx
HubSpot
Salesforce
REST

Ask the Database

Piloted with users

Questions in plain language, answers straight from the data. The AI only sees the database structure and writes the query; the query itself runs locally. Records never leave your system.

Natural language
Schema only
Runs locally

Core capabilities

Schema and database design
Data classification and enrichment
AI integration on top of real data
Long-term system maintenance

Approach

Founder-led, from the first conversation to production

datrmine is led by Ralph Ulrich. You work directly with the person who designs and builds the system, with no account managers or handoffs in between.

40+ years change the way you look at complexity. Technologies come and go; none of them is the goal.

Replace only for a reason.
Reliable systems stay in place unless there is a reason to replace them.
Integrate around reality.
Architecture follows the operation, not the other way around.
Handle the edge cases.
That is where operational reality becomes visible.

In practice

Expected: Centralisation project with an enterprise platform and external consultants.

Delivered in weeks

One central data layer for tug data across technology, dispatch, crewing and offshore. Every department kept its tools.

Expected: Database and reporting project to import historical crew experience.

Delivered same day

In about 40 minutes the data was inspected, special cases handled and the report adapted.

Contact

Describe the problem. We'll tell you how small it can be.

A direct conversation with Ralph Ulrich, the engineer who will build it. No sales process.

practice: datrmine
led by: Ralph Ulrich
focus: B2B · data · AI
reach: International

Keep what works.Fix what doesn't.Build what's missing.

What looks like a project

Built for operations, proven in production

Not just where a tug is, but whether it can go.

Readiness on the map

Competency in one click

From plan to deck, in sync

Keep your phone system. Automate the work behind the call.

Joins your phone system

Workflows are configuration, not code

Done, or queued for a human