01 / Easy Diavgeia / Search & product engineering

Making public information searchable.

Greek public-sector decisions contain valuable information, often spread across PDFs with inconsistent or missing text. I built Easy Diavgeia to make that information easier to find and follow.

RoleFounder & Principal Software Engineer
Period2023–2025
75M+indexed documents
~380 msp95 search latency
1M+PDFs repaired or OCR-processed

Scale and performance figures from September 2026.

The problem

Make a large, uneven collection of public documents useful as a search product. That meant handling ingestion, text quality, Greek-language ranking and ongoing updates alongside the everyday requirements of a subscription service.

My contribution

I independently designed, built and operated the full-stack platform: document ingestion, search APIs, frontend, authentication, billing, alerts and production infrastructure.

Engineering decisions

Reliable ingestion

Built asynchronous pipelines with revision handling, bulk indexing and retries. Selective OCR and text repair made more than one million scanned or malformed PDFs searchable.

Search that reflects intent

Combined Greek full-text search with phrase and proximity boosts, recency weighting and query-aware snippets. Production search achieved approximately 380 ms p95 latency.

A complete product

Delivered subscription-aware APIs, Stripe billing, usage quotas and deduplicated email alerts. Owned deployment, infrastructure migration, automated snapshots, error monitoring and worker health checks.

The result

A working search service with roughly 1,000 registered accounts and more than 75 million indexed documents. The project brings together the engineering and operational decisions needed to own a product end to end.

Start a conversation

Have an interesting problem
to work on?