// Research & Insights
RGL-LIB · 01

Research & Insights.

A curated library, not a blog. Proven expertise from production engineering and conference stages, perspectives on responsible AI, and peer-reviewed research. Assembled in one place so the technical depth is visible, not implied.

// Proven Expertise

Proven expertise.

Bylined engineering work and named conference contributions from the founders, curated for what RGL builds today. Depth you can verify, not just claim.

Benchmark, GPUExternal

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.

// Key takeaway

L4 delivers roughly 2x the inference throughput of T4 across every tracker tested, but only because the encoder/decoder bottleneck shifts before raw compute does.

Originally on Medium, 2023. Adrian Apap, co-author.
Read on Medium →
Benchmark, EdgeExternal

Is the new NVIDIA Jetson AGX Orin really a game changer? We benchmarked it.

An early hands-on benchmark of NVIDIA's then-new Jetson AGX Orin Developer Kit against the previous-generation Jetson AGX Xavier, across inception_v4, VGG19, super-resolution, U-Net segmentation, pose estimation, YOLOv3-Tiny, and ResNet50. Run on early-access software using NVIDIA's jetson_benchmarks scripts, modified to support Orin.

// Key takeaway

Even on early-access software, Orin landed at roughly 2x to 4x the inference FPS of the Xavier across every model tested, at the same form factor and price band. A step-change in what fits at the edge, not an incremental refresh.

Originally on Medium, April 2022. Adrian Apap, benchmarking team contributor.
Read on Medium →
Conference TalkExternal

How to start a business on NVIDIA Omniverse

Natalia spoke at NVIDIA GTC 2023 on the practical side of building businesses on top of Omniverse, covering simulation, digital twins, and the productisation gap between Omniverse demos and real customer deployments. Joint session with NVIDIA, Reallusion, Evolver, Aireal, ipolog, and SyncTwin.

// Key takeaway

The interesting Omniverse work is not in the rendering. It is in the data plumbing between the simulation, the real-world sensors, and the operational system that consumes the output.

NVIDIA GTC Spring 2023, Session S52058. Natalia Mallia, speaker.
View LinkedIn post →
Digital Twin, OmniverseExternal

Seeing double: a digital twin with a simulative spin

A walkthrough of building a real-time digital twin of an Antwerp road intersection in NVIDIA Omniverse, with weather profiles and a day-night cycle driven by live atmospheric data fed in from a deployed sensor box. Covers the 3D asset pipeline, the custom connector that pulls sensor data into the simulation, and the trade-offs between geometric fidelity and runtime cost.

// Key takeaway

Visual fidelity is the easy half of a digital twin. The hard half is the connector that gets real sensor readings into the simulation fast enough that the twin reacts to actual conditions, not yesterday's data.

Originally on Medium, May 2022. Natalia Mallia, co-author.
Read on Medium →
// Features

Talks, podcasts, and appearances.

Selected public appearances from the founders, including conference sessions, keynotes, and podcast conversations on AI delivery, governance, and industry adoption.

Upcoming SeminarExternal

University of Malta DSP Seminar, LLMs That Lie Less

Upcoming Data Science Platform seminar at the University of Malta on engineering responsible language models for production, covering source adherence, guardrails, evaluation in the wild, and cost-reliability trade-offs.

Wednesday 29 April 2026, 12:00-13:00. Faculty of ICT, ICT Communications Lab.
View event details →
KeynoteExternal

Lovable Women's Day event, Malta (Date Art Cafe, March 8)

Women-focused event in Malta hosted with Malta Startup Space and Lovable, featuring Natalia as a speaker on building with AI.

SheBuilds on Lovable event, hosted at Date Art Cafe.
View event post →
KeynoteExternal

AI Collective keynote

Keynote appearance on applied AI delivery and trustworthy deployment practice, focused on converting strategy into concrete implementation plans.

AI Collective appearance (source post via Axon Park instagram page).
View Instagram post →
PanelExternal

Malta and the EU AI Act, panel on responsible AI adoption

Natalia joined a national panel discussion on the EU AI Act in Malta, focusing on ethical AI, transparency, and practical implementation in education and industry.

LinkedIn post, EU AI Act panel discussion.
View panel post →
ConferenceExternal

SiGMA, Data is the new gold

SiGMA panel on turning raw data into usable business infrastructure, with a focus on governance, quality, and production-readiness in real deployment environments.

SiGMA panel appearance (video recording).
Watch panel →
PanelExternal

Start-Up Festival Malta, Decisions by Numbers

Natalia joined the Startup Festival Malta panel on how startups can use data for smarter decision-making, with practical strategies and founder-focused examples.

Startup Festival Malta panel post, Decisions by Numbers.
View panel post →
PodcastExternal

The SHE Word, Women and Everything You Need to Know About AI

Panel discussion on AI literacy, gender bias in mainstream AI systems, and practical ways everyday users can evaluate AI claims without relying on hype.

The SHE Word, S6 EP02 (YouTube episode).
Watch episode →
Live demoExternal

Building the Metaverse — live coding with NVIDIA Omniverse

Live webinar demo building digital twins in NVIDIA Omniverse: a synthetic-data pipeline for license-plate recognition, office digital twin walkthrough, and from-scratch extension development (live Python in Omniverse Code).

Deltatre Innovation Lab webinar (SmartCow × NVIDIA Omniverse).
Watch session →
// Perspectives

Strategic takes on trustworthy AI.

Where trustworthy AI succeeds, where it fails, and what most LLM coverage gets wrong. Collected from the founders' conference talks and ongoing LinkedIn writing.

Ongoing WritingExternal

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.

// LinkedIn highlightsRead more on LinkedIn →
Curated on LinkedIn, ongoing.
// Academic Work

Peer-reviewed research.

Published academic work from both founders. Peer review for enterprise and public-sector buyers who need more than a portfolio.

Peer-Reviewed JournalExternal

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.

// Key takeaway

Explainable, one-shot computer vision can match or beat black-box deep-learning baselines on identification accuracy, while staying transparent enough to satisfy the EU trustworthy-AI principles cited in the paper.

Neural Computing and Applications, Springer, 2024. Open access. Adrian Apap, co-first author.
Read the paper →
Published ResearchExternal

Escape the (Virtual) Room: procedural content generation for VR

Natalia's bachelor's research, published in the University of Malta's THINK research magazine. Combines virtual-reality interaction design with procedural content generation, using a genetic algorithm to generate escape-room puzzles and layouts that human players read as deliberately authored, not algorithmic.

// Key takeaway

Genetic-algorithm puzzle layouts produced VR escape rooms that participants described as natural and consistent, closing the gap between procedural generation and authored design.

THINK Magazine, University of Malta, Faculty of ICT.
Read the article →