End-to-end product · Telegram + Web
F-Signal
Marketplace intelligence delivered through Telegram and a Mini App.
I designed and built the product end to end: product logic, system architecture, Python backend, data and analysis pipeline, authenticated API, Telegram delivery, WebApp, validation, and release boundaries.
What it does
F-Signal monitors fresh Kufar listings, normalizes product facts, compares prices with market baselines, performs AI-assisted risk analysis, matches opportunities to purchasing profiles, and delivers selected signals through Telegram and a Mini App.
System shape
- Listing ingestion
- Normalization
- Market baseline
- Structured analysis
- Scoring
- Profile matching
- Durable signal state
- Telegram and WebApp delivery
Key engineering decisions
A staged pipeline
Each transformation has an explicit responsibility instead of mixing collection, analysis, ranking, and delivery.
Shared durable signal state
Telegram and WebApp consumers read the same persisted product state.
Protected operational boundary
Authenticated product APIs and operator tooling remain separated from the public surface.
Structured delivery system
The project was developed through my own structured delivery system: documentation, decisions, validation gates, release and rollback discipline, and separation between public product surfaces and private operations.
Result and evidence
The Hmara-hosted production MVP reached an owner-provided pipeline snapshot dated 2026-07-17, reviewed for this case: 2,993 stored listings, 5,444 seen listing IDs, 2,494 evaluations, 3,385 AI analysis runs with 3,258 successful, 2,215 unique listings analyzed by AI, 370 fetch runs, 555 Telegram notifications, and 21 test files. These figures describe pipeline volume and engineering maturity, not paid traction, revenue, customer growth, or established demand.
Scope boundaries
- The implemented first vertical focuses on iPhone marketplace listings.
- The private source repository, operational details, provider configuration, and user data are not publication artifacts.
- No paid-customer, revenue, demand, or scale claim is made.