Skip to content

Project Charter — IngoShift

Author: Velat Isik · Created: 2026-07-28 · Status: Phase 0 complete

Problem

The Ingolstadt economic region (Region 10) is undergoing a structural shift: from combustion-era automotive manufacturing toward software-defined vehicles, batteries, electrification, data/AI, cybersecurity, and energy transition. Decision-makers — job seekers, SMEs, economic-development analysts, recruiters — currently rely on scattered PDFs, one-off spreadsheets, and anecdote. There is no reproducible, source-traceable view of what is actually changing.

Product

IngoShift is a regional intelligence platform: repeatable pipelines that collect legally usable public data, normalize and validate it, historize it, and publish decision-oriented views with full lineage. The public demo covers Region 10; region and themes are configuration, so the same system adapts to other DACH regions and industries.

Positioning claim the project must support:

Most researchers deliver a spreadsheet. This project delivers the documented pipeline that produces, validates, updates, and explains the spreadsheet.

Scope — three core modules, completely, before anything optional

  1. Jobs & Skills Intelligence — skill/occupation demand, trends, co-occurrence, employer hiring activity, with observed-vs-inferred labelling.
  2. Company & Automotive Transformation Monitor — historized company model, auditable entity resolution, snapshot-based change detection, transparent transformation indicators (no black-box score).
  3. Tender & Research Radar — procurement notices and public research activity classified against the configured themes, with deadline views and relationships.

Shared platform: raw → bronze → silver → gold layers, quality gate that blocks publication on critical failures, lineage on every metric, FastAPI, one Streamlit dashboard, Excel/CSV/Parquet exports, reproducible demo build.

Out of scope (core release)

Graph database, RAG assistant, forecasting, alerting, scheduled email — each only after core acceptance criteria pass and with a written value justification (Phase 7). No paid APIs in the default demo. No login-protected or scraped-against- terms sources. No personal-data enrichment beyond what business questions require.

Target users

  • A — Job seeker / student: which skills, roles, tools, industries are in demand locally.
  • B — SME owner / sales team: repeatable company, market, hiring, tender intelligence.
  • C — Economic-development analyst: evidence on transformation, clusters, dependencies.
  • D — Recruiter / training provider: requested skills, occupational trends, training gaps.

Every major feature must map to at least one of these users (see user-stories.md).

Success criteria

Defined bindingly in definition-of-done.md. Summary: clean-clone demo build; three core modules working end to end; critical quality failures stop publication; every dashboard metric traceable to curated records and source; synthetic data visibly labelled; documented and tested API; CI green; README claims match implemented reality exactly.

Constraints and principles

  1. Honesty before impressiveness — every result classifiable as real / derived / synthetic / curated / inferred / unavailable; no fabricated coverage, depth, or readiness claims.
  2. Legal and ethical collection — official and machine-readable sources first; robots/ToS reviewed before any collector; constraints in data-governance.md.
  3. Reproducibility — documented, deterministic setup from a clean clone.
  4. Decision utility — every dashboard page supports a named user decision.
  5. Modular scope discipline — three finished modules beat twelve half-finished ones.

Phase plan

Phase 0 discovery ✅ → 1 foundation → 2 core data platform → 3 Jobs & Skills → 4 Company Monitor → 5 Tender & Research Radar → 6 productization → 7 optional intelligence. Each phase ends with the four-review standard (engineering, data, business, honesty) documented in docs/reviews/.

Known starting conditions (honest)

  • History accumulates from first pipeline run; only official statistical time series provide real backfill. Early trend views will be labelled with their true coverage.
  • Posting-level job data has no official free API; the default demo proves the extraction pipeline on labelled fixtures while official aggregates provide the real regional signal (ADR-008).
  • The company universe starts as a curated, evidence-backed seed — documented as partial coverage, never presented as complete (ADR-009).