Practical AI · Amsterdam, NL

From raw data to working AI/ML, data and software systems running in production.

Strategy, architecture & implementation. One person.

An engineer who combines strategy, architecture and implementation in a single role. I build pipelines, document intelligence, internal tools and production-grade ML systems that run every day. From raw, messy input to a working operational product.

Services

05 areas

Five areas where I make the biggest difference. Most projects touch two or three. Rarely just one.

01 AI/ML systems From prototype to production. Models, inference, evaluation, monitoring and integration into existing processes. A system that keeps running after I leave.
02 Data platforms & pipelines Batch, streaming and event-driven pipelines for documents, sensor data, logs and operational systems. Databricks, Spark, ETL. Production-ready, not demo-ready.
03 Document & knowledge systems OCR, extraction, classification, search and review workflows for large document collections and technical archives. At 100k+ scale, for contracts, specs and drawings.
04 Internal tools & decision support Custom software, dashboards and workflow tools that help teams work faster and more consistently. From prototype to production. Built to last.
05 Technical strategy & delivery Architecture decisions, roadmap, stakeholder alignment and hands-on execution in the same person. Technical depth and board-level clarity. No translation layer.

What I actually build

systems

No abstract capabilities. These are the things I actually deliver.

AI/ML

AI/ML pipelines running in production. From raw input to evaluated output, with monitoring and fallback.

Document intelligence

Extraction and search workflows at scale for unstructured document collections and technical archives.

Data pipelines

Pipelines for sensor data, logs and events. Batch or streaming, reliable in production.

Internal tools

Internal tools for operations, review and decision support. Built for daily use, not for a demo.

Prototypes → production

Prototypes deliberately designed to grow up. Not throwaway code — a foundation.

Technical roadmaps

Architecture decisions and roadmaps that stay legible to management. And technically sound.

Approach

04 principles
01 · Production-first

Nothing works until it runs on real data.

Prototypes are cheap. Production systems are where things go wrong. I start from the production side and work backward. Not the other way around.

02 · One person

No translation layer between pitch and code.

What I promise in a meeting I build myself. That removes the noise between sales and delivery where most consultancy projects come undone.

03 · Explainability

Technical choices stay explainable.

Leadership, engineering and operations hear the same message in their own language. That speeds up decisions and makes building possible.

04 · Messy reality

Real data. No toy datasets.

Real-world data is incomplete, contradictory, scales badly. That's where I spend most of my time. And where most others get stuck.

Examples of work

references

OCR and document pipelines for large technical document sets: extraction, classification and search at scale.

Data platforms for operational and sensor-driven workflows, realtime and batch.

AI/ML applications where reliability and explainability matter more than a flashy demo.

Internal tools and decision-support systems for operations and knowledge workers.

Stakeholder work spanning technical depth to executive decision-making. In the same project.

Architecture and roadmap engagements for organisations serious about building AI/ML, not just evaluating it.

Stuck on a data or systems problem with no idea where to start?

Email bob@tallydigital.nl LinkedIn linkedin.com/in/bobvdheijden Location Amsterdam · Netherlands