— DeepTech AI Engineering · Sovereign AI · Malta EU

Built To Last.

Most AI is built fast. Ours is built to last. We are a DeepTech AI engineering company building sovereign AI systems for EU organisations, compliant by design with GDPR and the EU AI Act, and hosted entirely on your infrastructure. Whatever your industry, your data stays yours. No shortcuts. No compromises.

Explore products →Start the conversation
Platforms
BEHOLDR · GRIMOIR
Sovereign AI
Your data stays yours
Also
Custom builds
// Our Approach

Sovereign AI.
EU-only By Design.

Production AI built on a sovereign architecture: your documents, video, and models stay inside your perimeter. DeepTech engineering and EU compliance built in, with platforms that hold up under scrutiny. That is what RetroGradient Labs builds.

01

Sovereign
By Architecture

Your data never leaves the EU. We build on EU-based infrastructure and avoid third-party AI cloud services, so there are no GDPR grey areas and no exposure to foreign data laws. What you build with us stays inside EU law, always.

02

Compliance,
Engineered In

GDPR and EU AI Act compliance are built into everything we ship, not added at the end. Audit trails, access controls, and human review on critical outputs so your legal and risk teams can sign off with confidence.

03

DeepTech
Delivery

Built by engineers who have shipped production AI, not consultants running pilots. Products when they fit. Custom engineering when they do not. The same standards either way.

// What We Offer

Sovereign AI Platforms.
Plus custom engineering.

GRIMOIR · BEHOLDR · CUSTOMDEEPTECH · MALTA EU

// OUR LINEUP · 01 OF 03
// SOVEREIGN AI PRODUCT · VISION

<BEHOLDR/>

Real-time answers from
cameras you already own.

Connect the cameras you already run. BEHOLDR turns live video into footfall, flow, and incident alerts your team can act on now, not in tomorrow's report. EU-hosted, GDPR-ready, EU-only cloud path.

  • Footfall & heatmapsQuantify movement, density, and busy zones without manual counts or delayed reports
  • Real-time alertsPush incidents to the right people the moment they happen. Act in the minute, not the morning after
  • Sovereign deploymentVideo and inference stay within the EU. EU-only cloud path, GDPR-aligned by design
Explore BEHOLDR →
RetailSecuritySmart CitiesManufacturingLogistics
The rules changed. Most AI hasn't.
Yours should.
GDPR
Fines of up to 4% of global annual turnover for data protection failures
EU AI Act
Mandatory risk classification, documentation, and human oversight for high-risk AI systems
Reputational risk
A single breach or non-compliance finding can cost more than the technology ever saved
Built for this
Every system we ship is designed to hold up under regulatory scrutiny from day one
// Founders

Built by
Engineers First.

RetroGradient Labs was founded by two engineers who have spent nearly a decade building production AI for businesses that needed it to actually work. We started this company because the gap between AI promises and AI products keeps widening, and the only way to close it is to build properly from the ground up.

Natalia Mallia, Co-Founder · CEO
Natalia Mallia
// Co-Founder · CEO

Natalia is a Maltese AI engineer with eight years of experience building AI that actually ships, in factories, classrooms, and creative studios. She holds an MSc in Artificial Intelligence from the University of Malta, speaks at NVIDIA's flagship developer conference, and was recognised across three categories at Malta's Best Businesswoman Awards 2025. At RetroGradient Labs she leads strategy, client relationships, and product, with one rule: AI is only useful if it's trustworthy enough to put into production.

NVIDIA GTC Speaker
SiGMA Featured Speaker
Responsible & ethical AI, hands-on
Malta's Best Businesswoman Awards 2025 · 3 categories
Adrian Apap, Co-Founder · CTO
Adrian Apap
// Co-Founder · CTO

Adrian is a Maltese AI engineer with over half a decade of hands-on experience building and shipping AI systems at scale, from real-time camera analytics on edge devices to AI-powered tools and automated trading systems. He holds an MSc in Artificial Intelligence from the University of Malta and is a peer-reviewed AI researcher, co-first author of a 2024 paper in Springer's Neural Computing and Applications on explainable, adversarially robust computer vision. At RetroGradient Labs he leads the technical vision with a builder's mindset and a deep conviction that robust, production-grade AI is what separates great ideas from lasting products.

Peer-Reviewed AI Researcher
Edge AI & Jetson Specialist
MLOps & Production Systems
Quantitative Research Background
// Research & Insights

Research & Insights.

Not a blog. A body of work. Engineering depth, conference stages, and peer-reviewed research from the founders, in one place.

Benchmark, GPUProven Expertise

Leveraging NVIDIA data-center GPUs for AI inferencing

Benchmarking the NVIDIA A2, A10, A40, A100, L4, L40, and H100 across a real production video-inference pipeline (PeopleNet, FaceDetect, face-mask classification). Practical findings on tracker choice, interval values, and where decoders become the true bottleneck.

Originally on Medium, 2023. Adrian Apap, co-author.
Read on Medium →
Ongoing WritingPerspectives

Natalia on responsible AI, LLM realities, and what production looks like

Strategic takes on where AI hype meets engineering reality: LLM limitations, the gap between trustworthy AI and AI marketing, responsible deployment in regulated sectors, and what enterprises actually need before adopting generative systems.

Curated on LinkedIn, ongoing.
Read on LinkedIn →
Peer-Reviewed JournalAcademic Work

Explainable multi-layer COSFIRE filters robust to corruptions and boundary attack, with application to retina and palmprint biometrics

A learning-free, hierarchical computer-vision approach to biometric identification on retina and palmprint images. Achieves perfect classification on the VARIA and RIDB retina datasets and 97.54% accuracy on the IITD palmprint dataset, while remaining robust to decision-based black-box adversarial attacks and to partial matching at 80% image visibility.

Neural Computing and Applications, Springer, 2024. Open access. Adrian Apap, co-first author.
Read the paper →
Browse all research →
// Get In Touch

Start a
Conversation or Discovery.

Have a use case in mind or not sure where to start? Tell us what you are trying to solve. We open with a scoped discovery or pilot so you can validate fit. Low commitment. High signal. Response within 48 hours.

● No purchase obligation. Initial consultation included. We respond within 48 hours.