AI Product Engineering: vision, voice and agent productsAimen Houalef
Complete AI products, not a chat box wrapped around an API call.
Document extraction with vision models. Real time avatars holding live voice conversations. Generated video pipelines. Agent workflows that act on real business systems. I have shipped all four.
What I have built
SnapMedia takes an uploaded PDF, extracts structured data with GPT-4o Vision, puts it in front of a person to verify, then renders a motion animated video with generated narration through a Remotion pipeline.
SherpaNova is a cooking companion built on a real time Simli avatar with live video and voice, where a response two seconds late is already a wrong response.
How I think about AI products
The model is the easy part. What separates a demo from something you can put in front of a paying customer is everything around it: what happens when the model is confidently wrong, whether a person sees the output before it matters, what gets logged so you can explain a decision three months later, and whether behaviour stays stable when the provider ships an update.
Any pipeline that puts raw model output in front of an end user is one hallucination away from an incident. I build the verification step in from the start rather than adding it after something goes wrong.
Working method
Claude Code and Codex are part of my daily workflow and have been for three years. They make me fast. Reading every diff myself is what keeps the output worth shipping.
FAQs
Starting at$9,000
Duration6 weeks
Tags
Next.js
OpenAI API
Python
TypeScript
AI Engineer
AI Integration
API Integration
Backend Engineer
Fullstack Engineer
Machine Learning
Service provided by
Aimen Houalef proDubai - United Arab Emirates
- $5k+
- Earned
- 1
- Paid projects
- 5.00
- Rating
- 9
- Followers
AI Product Engineering: vision, voice and agent productsAimen Houalef
Starting at$9,000
Duration6 weeks
Tags
Next.js
OpenAI API
Python
TypeScript
AI Engineer
AI Integration
API Integration
Backend Engineer
Fullstack Engineer
Machine Learning
Complete AI products, not a chat box wrapped around an API call.
Document extraction with vision models. Real time avatars holding live voice conversations. Generated video pipelines. Agent workflows that act on real business systems. I have shipped all four.
What I have built
SnapMedia takes an uploaded PDF, extracts structured data with GPT-4o Vision, puts it in front of a person to verify, then renders a motion animated video with generated narration through a Remotion pipeline.
SherpaNova is a cooking companion built on a real time Simli avatar with live video and voice, where a response two seconds late is already a wrong response.
How I think about AI products
The model is the easy part. What separates a demo from something you can put in front of a paying customer is everything around it: what happens when the model is confidently wrong, whether a person sees the output before it matters, what gets logged so you can explain a decision three months later, and whether behaviour stays stable when the provider ships an update.
Any pipeline that puts raw model output in front of an end user is one hallucination away from an incident. I build the verification step in from the start rather than adding it after something goes wrong.
Working method
Claude Code and Codex are part of my daily workflow and have been for three years. They make me fast. Reading every diff myself is what keeps the output worth shipping.
FAQs
$9,000