Reproducible, forensically grounded research

AI makes research faster. We make sure it still holds up.

AI makes research and documentation faster — and makes it easier to fabricate a source, lift someone else’s work without attribution, or lose track of how a conclusion was reached. Research-Ready builds software and a methodology that reverses that: research as traceable as forensic evidence, following the discipline of the Electronic Discovery Reference Model (EDRM) — self-hosted, so no one else can erase that trail.

8 production pipelines
100% self-hosted
NL/EU under European control
The technical foundation

Self-hosted, because control is a precondition for accountability

Every pipeline runs on a self-hosted AI stack. Reproducible means something concrete here: someone who wasn’t part of the project can reconstruct how a result came about from the stored data and the log alone — without asking the team. Data, source code, and models stay inside the environment you control, even while the pipeline makes decisions on its own.

  • Every step — who did what, when, based on what — sits in a version-controlled, timestamped archive
  • Services run with least privilege and verify each other before exchanging data
  • Passwords and keys are stored encrypted, never in plain text
  • No vendor lock-in: models and providers are interchangeable

Eight pipelines, one approach

From citable research to business creation — each pipeline is a self-contained, adaptable production line with its own repository.

01 · uc1-research

Citable reproducible research

Takes a research question in plain language, searches the open web, and sets a three-agent team — researcher, analyst, writer — to work bundling findings with full citations. Every finding lands in a knowledge graph, so the next question builds on prior work. The output is a formatted report, committed to a version-controlled archive with a traceable trail.

02 · uc2-app-builder

App and serious games builder

Takes a functional spec or design brief and autonomously works through specification, architecture, code, and tests, with automated checks on every commit. A human review checkpoint remains available via a browser IDE, but is never a required step.

03 · uc3-chatbot-voice

Chatbot with voice

Conversational AI with voice in and out: local speech recognition with a cloud fallback, memory that recalls earlier conversations, and optional image generation on request. Deployable for customer contact or internal services where voice is the natural interaction.

04 · uc4-supply-chain

Supply chain visualization

Maps a supply chain backward from a product or company name: who supplies this component, who supplies that supplier, where raw materials originate. Where direct data is missing, AI infers the missing links. The result is a searchable supply graph with dashboards for chain analysts.

View all eight pipelines →

The method

Research treated as evidence, not as claims

Every pipeline follows a forensic discipline for research and documentation — inspired by the chain-of-custody rigor of the Electronic Discovery Reference Model (EDRM), and developed through research into reproducible research infrastructure at the Value Chain Hackers Lab, Windesheim.

01

A fixed workspace, upfront

Every project starts inside a fixed structure. Sources, assumptions, and scope are settled before any data is collected or analyzed.

02

Reasoning gets recorded, not just the result

Every significant interpretive step is written down as it happens — not reconstructed afterward from the team’s memory.

03

Nothing gets overwritten

Earlier versions are kept, so the path from first draft to final result can be followed step by step.

04

A reconstructability check before handover

Before a pipeline goes live, we test whether someone outside the team can follow the conclusions using only the stored data and the log.

That way a citation stays a citation, a source stays a source — and a conclusion can be traced back to the evidence behind it.

More on where this comes from →

Cost and getting started

Fixed infrastructure, no per-user bill

Compute is spread across a compact server setup — no data center required. Want language models running fully locally, with no external fallback at all? That can be added once the need arises.

  • Predictable infrastructure costs instead of a SaaS bill that grows with usage
  • No per-user or per-seat pricing
  • Shareable across multiple teams or departments within the same organization

Start small, see results fast

  1. Weeks 1–2 Intake. Pick an owner and one process or pipeline.
  2. Weeks 3–6 Setup. Connect the platform to existing systems and data.
  3. Weeks 7–10 Pilot live. Runs alongside the existing workflow, with human oversight.
  4. Weeks 11–13 Evaluation. On quality, time saved, and cost — then a decision on scaling up.

One process owner · A pilot budget · Access to one bounded process

No long-term contract. No vendor lock-in. Visible from day one.

Curious what a pipeline could do for your organization?

Book a short, no-obligation call, or send us a message directly.