Mohammed Rhouta

The client needed to conduct phone prospecting at scale.
I had previously worked on a similar project using Vapi and Retell, but the results were unsatisfactory:
- Excessive latency
- Transfers between two agents created silences that gave the AI away.

I therefore redesigned the architecture with a single agent handling the entire conversation—from the gatekeeper to the executive—without any tr

When the pharmacist wasn't available, the team found themselves stuck when it came to advising customers. The slightest doubt meant no answer, and the customer would leave empty-handed.

I implemented an AI system that answers employee questions in Arabic and French, just as the pharmacist would. The AI consults a comprehensive medical knowledge base and provides verified advice.

The challenge: i

The client wanted to send personalized prospecting videos at scale.
The problem: a motion designer charges €300 to €500 per video and takes 4 to 6 weeks to produce it.

I looked for an approach to generate 100 per day, each one unique, without blowing the budget.

For the voice, ElevenLabs was the gold standard but remained too expensive at scale. I found an open-source model of equivalent quality

The client wanted a system capable of analyzing financial markets and executing trades autonomously, with real money on the line.

The real challenge isn't technical: it's that AI can be highly confident while still being wrong. An 80% confidence signal can still result in a loss of money.

I built an agent that consults a 30-hour knowledge base via a RAG system with specialized financial embeddin

Every morning, a trader spends 2 to 4 hours reading economic news, checking institutional positioning, analyzing volatility, and trying to understand which way the market will move. In a hedge fund, this job is handled by analysts paid over $100,000 a year.

I automated this entire process using five AI agents, each specialized in a specific data source.

The real engineering challenge is that th