Due diligence / RAG-style profiles · Independent project
Estonian Company Intelligence Hub: Registry, Web Signals and Risk Scoring
Dmitri Derevjanko built the Estonian Company Intelligence Hub, a due-diligence workspace that combines registry data, web signals, explainable risk scoring, and PDF reports.
This is one of 10+ independent public projects by Dmitri Derevjanko. 100+ additional workflows were delivered professionally, mostly under confidentiality, and are not documented here.
Open live demo Back to portfolio
Problem
Checking an Estonian company often means a registry page, a website, and a spreadsheet. Signals do not share a schema, and AI summaries without sources are not usable for handoff.
Solution
One canonical company profile: identity, contacts, activity, employees, tax and financial indicators, annual reports, web enrichment, company-scoped AI chat, deterministic risk scores, and HTML/PDF export.
What Dmitri Derevjanko built
Dmitri designed the data model, FastAPI APIs, Teatmik scraping with proxy rotation and anti-bot handling, web enrichment, grounded summaries, risk heuristics, PDF reports, structured logging, and a hardened Docker deploy with self-heal checks.
Architecture
- A fast probe runs before full scrape so empty or blocked lookups fail cheaply.
- Registry and HTML extraction normalize into one profile with source metadata.
- Open-web snippets extend the profile beyond a single registry.
- Risk scoring is rule-based and explainable; LLM text is grounded in the collected profile.
- Chat and reports stay scoped to the selected company.
- Request IDs, AI health diagnostics, and deploy guards reduce silent failure after restarts.
Technologies
- Python
- FastAPI
- Playwright
- OpenAI API
- Docker
- GitHub Actions
- PDF reporting
Result
A single analyst screen for Estonian company lookup, 0–100 risk scoring, AI summary/chat, and PDF handoff.
Limitations
- Depends on public registry and web sources; incomplete filings mean incomplete profiles.
- Scrapers break when markup or anti-bot rules change; fallbacks reduce but do not eliminate gaps.
- Risk scores are heuristic decision support, not a credit rating or legal opinion.
Live demo and source
Live demo: https://intel.dmitriderevjanko.com/
GitHub: source is not public. The live demo is the public artifact for this project.
Portfolio case study: https://dmitriderevjanko.com/projects/company-intelligence-hub/