Building a cheaper, more predictable AI pipeline I built Job Radar to collect vacancies from DOU ...Building a cheaper, more predictable AI pipeline I built Job Radar to collect vacancies from DOU ...
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Building a cheaper, more predictable AI pipeline
I built Job Radar to collect vacancies from DOU and Djinni, evaluate their relevance, save the results to Google Sheets, and send a Telegram digest.
The first version called OpenAI for every matching vacancy. It worked—but duplicate links, irrelevant roles, and repeated runs were consuming tokens without adding value.
I added a deterministic layer before the model:
• hard filters for obvious non-fits • local pre-scoring • URL normalization and deduplication • cached model responses • retries only for technical failures
The result is a smaller and more reliable AI boundary: fewer unnecessary calls, more predictable costs, and easier debugging.
Back to feed
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started