You will get a web scraper that runs on a schedule and warns you when it breaks
Project details
Most scrapers do not fail loudly. They keep running, keep writing rows, and quietly return nothing useful because a class name changed. You find out weeks later, from the data.
So the thing I build checks itself. On the scheduled tiers every run is measured against what the last one produced: row count, empty fields, values that stopped changing. If a run does not look like a real run, it is refused rather than written, and you get told.
I have built and shipped 50 public scrapers on the Apify marketplace, and this is the lesson that cost me the most to learn.
I will also tell you before you pay if a site cannot be scraped reliably, or if the data you want is not actually on the page. That is a cheaper conversation than a refund.
So the thing I build checks itself. On the scheduled tiers every run is measured against what the last one produced: row count, empty fields, values that stopped changing. If a run does not look like a real run, it is refused rather than written, and you get told.
I have built and shipped 50 public scrapers on the Apify marketplace, and this is the lesson that cost me the most to learn.
I will also tell you before you pay if a site cannot be scraped reliably, or if the data you want is not actually on the page. That is a cheaper conversation than a refund.
Data Tool
PythonWhat's included
| Service Tiers |
Starter
$140
|
Standard
$580
|
Advanced
$1,600
|
|---|---|---|---|
| Delivery Time | 4 days | 10 days | 21 days |
Number of Pages Mined/Scraped | 500 | 5000 | 20000 |
Number of Sources Mined/Scraped | 1 | 3 | 8 |
Number of Revisions | 1 | 2 | 3 |
Optional add-ons
You can add these on the next page.
Additional Source Mined/Scraped
(+ 3 Days)
+$200
Additional Revision
+$50Frequently asked questions
About Daniel
AI Video & Data Pipelines That Fail Loudly | Python, FFmpeg
Yavne, Israel - 9:19 pm local time
Most "AI video" work stops at a script. Mine ends at a rendered file. I build the whole chain: transcription, content-aware cutting, 9:16 reframing, burned-in word-level captions, batch rendering. Hours of raw footage become publish-ready clips without an editor ever opening a timeline.
On the LLM side I ship agents that survive real users:
- An AI SMS receptionist on Cloudflare Workers running Llama 3.3 70B, holding full conversations in English and Hebrew and capturing structured leads.
- A consumer app that returns per-item calories and macros from one photo, using Mistral Small 3.1 24B vision on Workers AI, with free and paid tiers.
- A nightly data pipeline indexing 416,817 job postings across 9 hiring platforms, published and running on a schedule.
What I do well:
- Whisper + FFmpeg pipelines: content-aware cutting, 9:16 reframe, animated captions, batch rendering
- Production LLM agents and API integration: Cloudflare Workers AI, Llama, Mistral, Claude, MCP
- Large-scale scraping and data pipelines that keep running after handoff
- Full deploy and ownership: Cloudflare Workers, D1, Pages, iOS and Android release
Microsoft certified: MCSA and Azure.
If you have hours of raw footage, or an LLM prototype that needs to handle real traffic, send it over and I will tell you exactly how I would build it.
Steps for completing your project
After purchasing the project, send requirements so Daniel can start the project.
Delivery time starts when Daniel receives requirements from you.
Daniel works on your project following the steps below.
Revisions may occur after the delivery date.
One page proved before the rest
I scrape a single page first and send you the row it produces. You confirm the fields are right before I run it at scale, which is when mistakes get expensive.
Full run, gated and handed over
The full run goes out, with the health checks and your schedule wired in if you bought them. You get the data, the code, and a short note on what breaks it.