melwyn.ai
← Lab / Selected work
Procurement intelligence

Tender Monitor

Turn hundreds of irrelevant listings a day into a short, ranked shortlist.

Built in-house for a business development team
The problem

Relevant opportunities were scattered across about seven incompatible procurement portals: different tech stacks, some behind logins, some with broken security, each publishing hundreds of irrelevant listings a day. Checking them all by hand was impossible, and a plain keyword search produced far too much noise to act on.

The approach

One workflow fans out to every source in parallel, dedupes and threshold-filters the firehose, then lets a fast AI model score only the survivors for true relevance against the team's criteria. The result is a short, ranked email. The same shape runs across three countries, a sibling variant inverts the filter to catch the training and events listings the main monitor throws away, and a zero-result email always sends so a silent failure never hides as a quiet day.

How it runs

Walk the pipeline

Click through each stage to see what the system does and what the team gets back.

Stage 1 of 5

Fan out to every portal at once

It pulls from about seven incompatible portals in parallel, whatever their stack: plain HTML, REST and GraphQL APIs, and JavaScript apps driven through a headless browser.

7 portalsin parallel
What changed

7

incompatible portals unified

3 countries

in one daily digest

Hundreds → a handful

that actually fit

Checking the portals by hand was never going to happen, and keyword alerts buried the real opportunities in noise. Now the team opens one ranked email a day and sees only the listings worth a bid. The same monitor runs across three countries, and because it sends even on empty days, a broken portal shows up as a missing result instead of a silent gap.

Parallel scrapersHeadless browserAI relevance scoringEmail digest

Want something like this on your team?