ROTOP Data Lake (project)

Delivery of ROTOP’s data lake and Claude Cowork layer, active since the tech kickoff on 2026-06-26. Grew out of the won ROTOP Data Lake deal. Bianca Frost is the Gemma project lead; Bijan ran the sales phase and the first working session.

Scope

A breadth-first MVP across verticals (Sales, Finance, Production, CDMO), built on a static Phase 1 snapshot: a one-time Business Central + Sage dump on a ROTOP Linux server, staged and cleansed into Snowflake, with a semantic/metadata layer that Claude Cowork reads. The flagship demo is consolidation across ROTOP’s two legal entities (Pharmak + Pharmazie). Explicitly not a BI/dashboard build. Phase 2 (a recurring pipeline) waits on ROTOP’s carve-out from the MDG group (earliest end of September 2026).

Stack

Loading with dlt plus Airflow on the ROTOP Linux server (no Fivetran/Airbyte). Snowflake (EU/Frankfurt) as the warehouse/data lake. dbt Cloud (Starter) for transformation, with its semantic layer exposed to Claude Cowork over the MCP server (no separate Airflow server for dbt). GitHub for version control. No classic dashboards: Claude Cowork is the primary interface. ROTOP-hosted Linux server (Strato); Gemma access via SSH plus IP whitelisting.

Status (as of the 2026-06-26 kickoff)

• Data: Business Central dump received (8.6 GB zip / ~19 GB unzipped) via SharePoint. Sage dump to follow (Frank Gaebler, back the following week). Bijan to review the BC data and flag anything unexpected.
• Access: Linux server at Strato; Bijan’s SSH key loaded; Stefan Profus to send the server IP; Bijan then adds Bianca’s key. IP whitelisting deferred to the first weekly sync.
• Accounts to set up: Snowflake trial (Gemma; ROTOP to register a credit card before Rex’s holiday in ~2 weeks), dbt Cloud trial (same), GitHub org + three repos (infra/IaC, data loading, dbt transformation; Frank Gaebler).
• Governance: Weisungsberechtigung to be formally delegated from Peterseim to Rex per the AVV (email mechanism agreed). Cloud Cowork admin/settings access for Gemma to be sorted within 14 days (DSGVO org settings).
• Cadence: weekly sync Tuesdays 13:00 (Bianca to send the invite; Bianca posts meeting notes into the ROTOP Teams channel). Gemma shares credentials via 1Password.

Plan and effort

Internal estimate (2026-07-01): ~22 days, ETA mid-August. Split: infrastructure setup 2.0 + data ingestion 2.5 (Bijan); dbt base 2.5 + dbt transformation 5.5 + MetricFlow semantic layer 4.0 + Claude Cowork integration 1.0 (Christophe); project management + workshops 2.5 (Bianca); 2-day buffer. After the kickoff, delivery moves into requirements engineering for the Finance metrics (field definitions + granularity, mirroring the Sales “Annex A”), aligned with Peterseim.

Team

Bianca Frost (project lead), Bijan, and Christophe Guillonnet, the primary engineer (carries ~13 of the 22 days). Working language is German (Christophe: written German, spoken English).

Delivery-phase meetings live under this project (see the Meetings collection), starting with the tech kickoff.