Data Lake Technical

A 40-minute technical 1:1 via Google Meet, Bijan and IT lead Torsten Rex, following the morning kickoff. No other stakeholders; focus on data access, infrastructure, stack, and next steps.

MDG carve-out (critical context)

ROTOP is a tenant inside the shared MDG group infrastructure: several subsidiaries sit in one Business Central and one Sage database, separated at the mandate/UI level. ROTOP is carving out into its own infrastructure, with parallel workstreams (Microsoft tenancy migration, network/firewall, and the BC carve-out, either mid-July to September or split off later). Earliest realistic exit: end of September. Consequence: Gemma cannot get a direct read-only DB connection (MDG would have to agree, and Rex expects they will not), and a daily dump is also uncertain.

Agreed data-access approach

  • Phase 1 (unblocks work now): a one-time BC + Sage dump, frozen ~end of May, ~6 months history, hosted on a ROTOP Linux server; Gemma connects via SSH (public-key exchange) + IP whitelisting from the Gemma VPN. ROTOP data stays on ROTOP infrastructure throughout.
  • Phase 2 (post carve-out, ~Q4 2026): a proper recurring/incremental pipeline once ROTOP has its own DB.

Rationale: build and demo on a static snapshot (mockups for Peterseim and Gottlieb) without pressuring MDG for a recurring feed.

Infrastructure and stack

Rex to provision a Linux box (on-prem or any cloud; Gemma works with anything). Access via SSH key + IP whitelist. GitHub agreed for version control (ROTOP has barely-used Azure DevOps). Stack: Snowflake warehouse (better access management + MCP integration than Postgres), dbt Cloud paid (built-in MCP server + Claude Cowork connectivity), GitHub.

ROTOP Pharmak (non-radioactive) and ROTOP Pharmazie (radioactive, incl. theranostics). Sage is used only in Pharmazie; BC covers both. Consolidating the two is exactly Peterseim’s Phase 1 demo case. Minimum viable lakehouse: BC + Sage, staging, cleansing, metadata layer, two consolidated entities.

Data volumes and follow-ups

Volumes unknown at the call (the DB admin was on holiday); Rex to confirm by 2026-05-27 (later reported BC ~20 GB, Sage ~50 GB). Rex asked Bijan to write a technical setup document (preferred option + an Option B fallback) to circulate internally; this becomes the technical annex to the offer. Commercial structure: NDA signed, formal engagement to come via Colayer once Gemma’s offer is wrapped. Rex floated a future upsell (Bijan presenting to ROTOP’s internal AI forum ahead of FY2027 budgeting).