I fix stalled AI projects,
build new ones from zero,
and scale what works.

Mazan Technologies is a specialist AI/ML practice led by Mehdi Salmani, Ph.D. — with senior practitioners brought in when the engagement requires scale. Not a staffing agency. Not a strategy firm. Hands-on technical leadership for enterprises that need AI systems in production, not in PowerPoint.

Recent work: asset-health analytics for 14,000+ airport HVAC assets · multi-agent orchestration for industrial workflows · production AI systems deployed across 130+ QSR locations · AI venture validation across aviation, automotive, and energy
Services
01

Fix

Your AI project stalled between POC and production. I diagnose why — data pipeline, model architecture, evaluation gaps, team structure — and get it shipping. Typical engagement: 4–8 weeks to production readiness.

02

Build

You need a production ML system built from zero. LLM pipelines, multi-agent orchestration, RAG architectures, real-time inference. Concept to deployed in 8–12 weeks, not 8–12 months.

03

Scale

Your AI works in one division. Now it needs to work across the enterprise. I build the platform architecture, the evaluation framework, and the operational muscle to get there — then hand it off to your team.

Selected Work
Selected ventures and production systems.

EverGuard.ai

Computer vision safety system for a Korean industrial giant. Reduced safety incidents on factory floors. Spun out as an independent company.

Spinout

Cyclope.ai

Deep learning vehicle detection system. Saved the client $5M annually in operational costs. Spun into a new business unit.

$5M annual savings

Gaido AI

Health deterioration detection models supporting real-time patient monitoring in clinical settings.

Acquired

Pull Systems

Battery lifetime prediction system for a German automotive OEM. Spun out as an independent venture.

Spinout
Background
Mazan Technologies
2025 – Present
Enterprise AI consulting. Fortune 100 clients. Multi-agent systems, LLM pipelines, venture validation across aviation, energy, automotive, manufacturing.
IBM Watson Orders
2020 – 2025
Head of Data & Analytics. Led 17 engineers. Scaled voice AI from 1 store pilot to 90%+ automation across 130+ stores. Built the data platform, annotation system, and evaluation infrastructure.
BCG GAMMA → BCG DV
2017 – 2020
Lead Data Scientist. Built 15+ venture-scale AI products for Fortune 500 clients. Three acquisitions or spinouts. Computer vision, NLP, fraud detection, signal processing.
SanDisk / Western Digital
2015 – 2017
Senior Data Scientist. Company-wide award for Spark-based data infrastructure used by SanDisk and Toshiba.
AI Startup (Co-Founder & CTO)
2003 – 2008
Built first Persian OCR engine. Scaled from 5 to 40 employees. $400K seed funding. Shipped B2B/B2C AI products.
Industries
Healthcare Manufacturing Insurance Automotive Aviation Energy Retail & QSR Logistics Pharma Public Sector
About
Mehdi Salmani

I'm Mehdi Salmani. I've spent 20+ years building production AI systems — at IBM Watson, BCG, and now through my own practice. I started Mazan Technologies because the hardest problems in enterprise AI aren't solved by another strategy deck. They're solved by someone who's built the system, debugged it at 2am, and shipped it under pressure.

When an engagement requires more than one person, I bring in senior practitioners from a vetted network of ex-BCG, ex-IBM, and independent AI/ML specialists. No juniors. No bench warmers. Only people I'd trust with my own clients.

Education
Ph.D., ECE — Purdue University
Research
7 journals · 20+ conferences · 3 Best Paper Awards
Citizenship
United States
Base
Los Angeles, CA
"The bottleneck in enterprise AI is never the model. It's the last mile between the demo and the system your team actually trusts."

Let's talk about what's stuck.

Typical engagements run 4–12 weeks for enterprise teams with a real deployment problem. Not exploratory AI brainstorming.

Discuss a Stalled Project
mehdi.salmani@mazantech.com (213) 598-9346 LinkedIn X