Anonymized case studies

Credible systems, described without exposing client names.

Dragonflow anonymizes client work on the public site. These examples describe the type of operational problems solved, the system built, and the business outcome.

Selected work

Operations work that becomes infrastructure.

Multi-location restaurant group - POS data automation

Replaced fragile local scripts with a cloud automation that logs into a legacy POS daily, extracts sales, cost, and inventory reports, and loads them into a ClickHouse warehouse. The system computes theoretical ingredient consumption daily.

Eliminated manual report pulls.
Improved visibility into inventory discrepancies.
Moved a local workflow into monitored cloud infrastructure.

Government-procurement intelligence product

Built a pipeline that turns raw public procurement data into monthly structured signals used by a newsletter and dashboard.

Normalized raw CSV and scraped data.
Produced recurring intelligence outputs.
Created a repeatable path from source data to product.

Regional market-list build

Delivered a one-time verified-business dataset of roughly 1,000 to 2,000 contacts with verified physical addresses, using tiered enrichment to stay within approved API limits.

Clean, spec-compliant data deliverable.
Physical-address verification included.
Repeatable process for new regions.

AI experience-feedback platform

Designed AI-moderated feedback loops embedded into SaaS and e-commerce products at the moments that matter, replacing static survey forms with adaptive agents.

Faster user-research loops.
Adaptive questions based on session context.
Less developer time needed to collect product feedback.

Technology stack

A practical stack for automated operations.

Next.js Supabase (Postgres + RLS) Tinybird / ClickHouse Render Playwright Claude / LLMs Google Places