Andre Hassan Wilson

1. Kenya property data platform and pricing index

Problem. Kenya has no public register of property sale prices, so buyers, especially those abroad, cannot tell whether a price is fair.

What I built. A pipeline that scrapes and ingests listings, normalises them into canonical properties, removes duplicates, enriches them with spatial data, and publishes an asking-price index.

• Data model: raw ingest, listing, canonical property with H3 hexagon indexing and recorded match decisions, in a PostgreSQL database of 114 models.
• Deduplication: 645 listings processed, 93 flagged as duplicates.
• Spatial enrichment: 2,199 geographic features and 407 area signals across 84 places, from OpenStreetMap, Google Open Buildings and regulator sources.
• Public index of 422 asking prices from 6 sources, labelled as asking prices, not sale prices.
• Licence matching against 703 records from the estate-agent register (Kenya Gazette).
• Ten scheduled jobs keep the data fresh.

2. Asset grading and market-check engine

Problem. Investors need a consistent, explainable view of whether a property and its price stack up.

What I built. A deterministic five-dimension weighted grading model and a market check that compares an asking price with the local median and its confidence interval.

• Example output: a Westlands asking-price median of KES 10.1M across 83 listings, with the subject price placed against the interval.
• Same inputs always give the same grade, so every score can be traced back to its components.
• Fed a product rate card and API pricing tiers.

3. Private equity document registry and knowledge graph

Problem. Partners across four countries worked from scattered files, chats and emails, with no single source of truth for deals, organisations and decisions.

What I built. A typed registry that classifies every document and extracts entities into a cross-referenced catalog, with automated quality checks.

• 552 documents catalogued and 270 entities typed across seven classes (deals, organisations, people, licences, events, topics, vehicles).
• Continuous integration rejects any change where the stored registry does not match a fresh rebuild.
• Retrieval accuracy test: 19 of 25 on the first run, 25 of 25 after fixes, with a clean rebuild changing zero files.

4. Weekly multi-source data pipeline

What I built. A scheduled pipeline that runs every Monday at 06:00. It pulls new material from Google Drive, Gmail and messaging exports, loads it into the document set, rebuilds the registry, runs validation and commits the result.

• Sources at build time: 160 Drive documents, 35 email threads and 3 group chats.
• Incremental loads with OAuth-authenticated Google APIs and a full validation pass on every run.

5. Sanctions and politically exposed person screening

What I did. Screened deal counterparties against government sanctions lists in three rounds, covering more than 55 named subjects.

• Sources: US OFAC SDN and consolidated lists, UK OFSI, EU financial sanctions file and UN list, processed as bulk CSV and XML.
• Each screen records the date of every list used and every source that failed, so the coverage limits are visible to decision-makers.

6. Critical-minerals market research and pipeline screen

What I did. Produced the partnership's market overview, a screen of the project pipeline against investment criteria, a capital-markets analysis and a register of buy-side demand. Also rebuilt financial models and benchmarked margins for proposed business lines in Excel.

7. Investor portal with KPI reporting

What I built. An investor portal with data-driven KPI tiles, a capital-account view, a deal-vetting assistant and a partner task board with digest emails.

• PostgreSQL with row-level security and tiered access, so each investor sees only their own data.
• Mandatory time-based two-factor authentication for every portal user.

8. Live CRM revenue dashboard and attribution

What I built. A revenue dashboard reading live from the Zoho CRM API, with lifecycle, pipeline, lead-source, affiliate and unit-inventory views, plus a native affiliate attribution model.

• Cleaned the lead base of duplicates and invalid records before reporting on it.
• 180-day first-touch attribution with automated CRM workflows and CSV export.
• Tracked availability and status across a residential development.

9. Frontier markets data-source catalog and search

What I built. A searchable catalog of data sources for Africa and Central Asia with country pages and answers that must cite their sources.

• About 185 sources loaded through a catalog ETL step into PostgreSQL with full-text and trigram search.
• 208 automated tests and passing acceptance tests.

10. Finance ledger and KPI scoreboard

What I built. A double-entry ledger with defined burn and runway metrics and positions kept separate by currency, plus a scoreboard where schedule variance is calculated from overdue milestones rather than typed in.

• Burn is defined as recurring commitments plus trailing 90-day spend divided by three, so the number cannot be edited by hand.
• A read-only status script reports pipeline KPIs straight from the database.

11. Development feasibility and unit-economics models

• Townhouse land appraisal, Nairobi: residual land valuation with price and unit-size sensitivity. Includes a base case and the breakeven sale price per unit.
• Phased apartment financing: modelled the gap between buyer instalments and contractor payments to size borrowing to the peak gap instead of the full build cost.
• Café concept, Nairobi: unit economics compared against the cost of debt, which led to a recommendation not to proceed at that scale.