Your organization is adopting AI faster than it understands what can go wrong.

AI hallucinates. It deceives. It produces behaviors even its creators cannot predict. When people trust it blindly, the cost compounds in silence — until it becomes a crisis.

We audit what technical audits miss: the gap between what your AI does and what your organization thinks it does.

See How We Help Book an AI Review Briefing

Dheiver Santos speaking on stage at an AI event

Dheiver Santos presenting Mangaba.AI at a tech conference

Dheiver Santos, PhD, speaking at the Summit Inteligencia Artificial

The standards behind this practice, forged at:

SX Negócios

The Blind Spot No One Is Talking About

Your AI deployment looks successful. The real question is: what happens when it fails?

Most organizations measure AI by speed and efficiency. But the ugliest failures don't come from models that underperform. They come from models that overperform — in ways no one expected.

Hallucinations that look like truth. Emergent behaviors no one coded. Delegation patterns where humans stop questioning the machine. These are not edge cases. They are the new normal.

We don't offer compliance checklists. We map the specific, prioritized, and actionable risks at the intersection of human behavior and AI behavior — where conventional audits stop looking.

See how we address it

How We Work

Three ways to close the gap between what your organization thinks its AI is doing — and what it actually is.

Hands-on executive training session with a laptop

01. Executive Training & Programs

For leaders making decisions with AI, not just about AI.

Most executives know how to use AI. Few understand where it fails and how that failure shapes the decisions they make every day.

This is not a workshop on prompting. It is a structured program on hallucination, emergent behavior, systemic bias, and responsible delegation, built for leadership teams that need to ask better questions, not just move faster.

Delivered as: In-person or remote sessions, tailored to your sector and your organization's actual AI use cases. Custom programs for boards, C-suite, and functional teams.

Book Executive Training

Desk with business charts, laptop and documents under review

02. AI Review & Assessment

What is your organization trusting AI to do, and what happens when it is wrong?

We examine the model, the architecture, the process, the strategy, or the costs. We identify where human judgment is being replaced irresponsibly, and surface the behavioral and regulatory risks that technical audits miss.

The deliverable is not a compliance scorecard. It is a real risk map: specific, prioritized, and actionable.

Delivered as: Structured assessment through interviews, process review, and documentation analysis. No code access required. Written report, presentation, and follow-up session included.

Also available as: Advisory board participation and independent expert sessions for organizations that need a critical external voice on an ongoing basis.

Request an AI Review

GPU server cluster used for training and fine-tuning custom AI models

03. Custom AI Model Training & Fine-Tuning

For companies that want their own model, not a rented one.

Off-the-shelf APIs make every company sound the same, and route your data through infrastructure you don't control. We fine-tune and train language models on your organization's own data, vocabulary, and workflows.

The result is a proprietary model you own end to end: deployed on your own infrastructure, independent of Big Tech pricing, rate limits, and policy changes.

Delivered as: Data preparation, fine-tuning (LoRA or full) and evaluation, plus deployment guidance for on-premise or private-cloud infrastructure. Includes a hands-on training track for your technical team.

Talk Fine-Tuning

Why Dheiver Santos:
Independent clarity in a world of vendor promises.

Every AI vendor promises speed and efficiency. But who tells you what your AI is actually doing — and what happens when it gets it wrong?

We have no software to sell. No platform to license. No 18-month engagement to pitch. We are independent advisors, which means our only incentive is an honest, accurate picture of your organization's real exposure.

The organizations that navigate this moment well are not the ones that adopted AI first. They are the ones that adopted it with the clearest understanding of what they were actually doing.

This practice is led by Dheiver Santos, PhD — machine learning engineer and AI researcher, with a PhD in Chemical Engineering with emphasis on Artificial Intelligence (UFBA/UFRJ) and postdoctoral research at UNICAMP/UNIT (CNPq) and UFAL (FUNDEPES).

His background is not theoretical. It was forged in senior industry roles — Senior Staff Machine Learning Engineer at SX Negócios and Senior Scientist at Grupo Boticário (GAVB) — and in the classroom, as a professor of Artificial Intelligence, distributed systems, and data science at UPE and Estácio.

He is the creator of Mangaba.AI, an open-source framework for teams of autonomous AI agents, and founder of Cognai, a healthtech applying AI to cardiology. His academic work spans machine learning and neuro-symbolic computing, with 550+ citations and an h-index of 12.

Speaking & Keynotes

Recognized voice on AI risk, emergent behavior, and responsible AI adoption at industry conferences and corporate events.

Summit Inteligência Artificial

Keynote on AI hallucination, emergent risk, and why the human-model interface is the new audit frontier.

Dheiver Santos presenting Mangaba.AI at Latinoware 2026

Latinoware 2026 — Mangaba.AI Launch

Presenting autonomous AI agent frameworks at the 23rd Latin American Congress of Free Software and Open Technologies. See profile ↗

Book Dheiver for Your Next Event

0 Academic Papers

0 h-index

0 AI Products Built

0 Patent (INPI)

0 Years Research & Industry

Google Scholar ↗ GitHub ↗

Products Built

Applied computer vision and machine learning shipped into real clinical and operational workflows.

AI model segmenting and tracking surgical instruments on a tray

Surgical Instrument Detection

Real-time segmentation tracks every instrument on the surgical tray, reducing the risk of retained items.

AI pose-estimation model monitoring a patient's posture in a clinical room

Patient Fall & Posture Monitoring

Pose-estimation model watches patient position around the clock and flags falls or unsafe movement instantly.

AI model segmenting individual teeth on a panoramic dental X-ray

Dental X-Ray Analysis

Automated tooth-by-tooth segmentation on panoramic X-rays speeds up diagnosis and treatment planning.

AI model detecting vitiligo on skin with a confidence score

Skin Condition Detection

Dermatology model identifies conditions like vitiligo directly from a photo, with a confidence score for triage.

AI model detecting and counting white blood cells under a microscope

Blood Cell Counting

Microscopy image analysis detects and counts white blood cells automatically, accelerating lab screening.

Frequently Asked Questions

What does an independent AI advisor do?

An independent AI advisor evaluates how your organization uses AI — models, processes, and human oversight — without selling software or platforms. The only deliverable is an accurate picture of your actual exposure, plus training so your team can act on it.

What is corporate AI model fine-tuning?

Fine-tuning adapts a language model to your company's own data, vocabulary and workflows, producing a proprietary model you can run on your own infrastructure — independent of Big Tech pricing, rate limits and policy changes. We use LoRA or full fine-tuning depending on the case.

How does an AI Review & Assessment work?

It is a structured assessment through interviews, process review and documentation analysis — no code access required. The deliverable is a specific, prioritized and actionable risk map, with a written report, presentation and follow-up session.

Who is Dheiver Santos?

Dheiver Santos, PhD, is a machine learning engineer and AI researcher with a PhD in Chemical Engineering with emphasis on AI (UFBA/UFRJ), creator of the open-source framework Mangaba.AI and founder of the healthtech Cognai. He has 550+ academic citations and industry experience at SX Negócios and Grupo Boticário.

Don't wait for the blind spot to become visible to everyone else.

Training & Programs · AI Review & Assessment

dheiver.santos@mangaba.ia.br

Or reach out directly. We respond within 24 hours.

Chat on WhatsApp