You will get Claude Code hooks that block what your written rules only ask for

Let a pro handle the details

Buy Other AI & Machine Learning services from Daniel, priced and ready to go.

Let a pro handle the details

Buy Other AI & Machine Learning services from Daniel, priced and ready to go.

Project details

A rule written in prose is a request. Claude can read it, agree with it, and still not follow it, because nothing in the system is checking. A hook is the same rule with teeth: it runs, it returns a verdict, and a bad action does not happen.

Two things I do differently, and both come from getting them wrong first.

The rule is inverted. A blocklist of dangerous verbs always loses, because the next spelling is not on the list. My guards deny anything touching a protected path unless every verb in the command is read-only, so new tools and new phrasings fail closed instead of open.

Every gate is proved by breaking it. I neuter the guard on purpose and show you the test suite going red, then restore it and show it green. A suite that stays green when you delete the thing it guards is decoration, and I would rather you see that demonstrated than take my word for it.
AI Development Type
Software Maintenance
What's included
Service Tiers Starter
$95
Standard
$450
Advanced
$1,200
Delivery Time 3 days 7 days 14 days
Number of Revisions
123
AI Model Integration
-
-
-
Detailed Code Comments
-
Knowledge Graph
-
-
-
Model Documentation
-
-
Ontology
-
-
-
Source Code
Taxonomy
-
-
-
Optional add-ons You can add these on the next page.
Additional Revision
+$60

Frequently asked questions

Daniel M.Status: Offline
Daniel M.Status: Offline
AI Video & Data Pipelines That Fail Loudly | Python, FFmpeg
Yavne, Israel - 4:38 am local time
I build AI systems that run unattended in production. Not demos. Shipped systems.

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.

The failing case, reproduced first

Before writing a gate I reproduce the thing you want stopped, so we both agree on what the hook has to catch. If I cannot reproduce it, I say so rather than guess.

Gate delivered, then broken on camera

You get the hook, its config and its test. Then I delete the guard so you watch the suite go red, and restore it so you watch it go green.

Review the work, release payment, and leave feedback to Daniel.