The Founder

Yousef Hossni

Founder & Lead AI ArchitectNeusora

Yousef designs Neusora’s intelligence end to end — from the retrieval pipelines and the privacy layer to the verification behind every answer.

He architected and deployed both of Neusora’s production models, N1 and Veritas. He is also a published AI researcher: his work on deep learning for wearable-sensor and multimodal data appears in Scientific Reports, where he is corresponding author, and in Springer Nature proceedings as first author.

LinkedIn↗ Google Scholar↗ ResearchGate↗ yousef@neusora.com

Research taught me that a claim is only as strong as its evidence. Every Neusora model is built to that standard: grounded, verified and private.

Research & Recognition

Scientific Reports2025

56+

Citations

Corresponding author

Springer Nature2026

99.78%

Accuracy and F1-score

First author · Opportunity benchmark

AI HackathonSolo entry

1st

Prize

Against 20+ competing teams

NeusoraLive

Production AI models

N1 · Veritas · architected end to end

Jan 2024Research beginsDeep learning for wearable-sensor and multimodal data, from his second year of engineering.
Feb 20241st PrizeAl Mulla Exchange AI Hackathon, as a solo entrant against 20+ teams.
Nov 2024Embassy R&DAn edge-AI prototype, co-presented to the Hungarian Ambassador.
Jan 2025Scientific ReportsCorresponding author of a KFAS-funded study on attention-based deep learning for activity recognition.
May 2025Neusora foundedThe same month, presents his research at the S4IoT’25 symposium.
Feb 2026Springer NatureFirst author: a multi-head attention CNN reaching 99.78% on the Opportunity benchmark.
TodayN1 and Veritas liveBoth production models, architected end to end. Engineering degree completed in 2026.

Publications

Peer-reviewed,
and verifiable.

Yousef’s published research applies attention-based deep learning to wearable and multimodal sensor data, recognizing human activity with up to 99.78% accuracy. Each paper links to its permanent record.

2026

Springer Nature · S4IoT’25 Proceedings · First author · Oral presentation

Activity Recognition Using a Multi-head Convolution Neural Network Integrated with Attention Mechanism

Yousef Hossni and Murad Khan. A four-pathway convolutional network with attention reaches 99.78% test accuracy and F1-score on the Opportunity multimodal sensor benchmark. Presented in person at the IEEE–Springer S4IoT’25 symposium, May 2025. Lecture Notes in Electrical Engineering, vol. 1513.

DOI ↗
2025

Scientific Reports · Nature Portfolio · Corresponding author

A comparative analysis of LSTM models aided with attention and squeeze and excitation blocks for activity recognition

Murad Khan and Yousef Hossni. Scientific Reports 15, 3858 (2025). Yousef ran every simulation and analysis in the study and built its deep-learning pipelines. Its open-access publication was funded on merit by the Kuwait Foundation for the Advancement of Sciences, and the audited simulation code is public. Cited 56+ times.

DOI ↗Code ↗

What He Built

Every layer of Neusora,
architected end to end.

Neusora’s two production models, N1 and Veritas, run on an architecture Yousef designed and built himself — from the first document ingested to the last answer delivered.

01 · Knowledge

Retrieval at scale

Pipelines that turn large volumes of unstructured material — PDFs, Word files, images, audio, video and scanned pages — into embedded, searchable knowledge, now scaling to millions of retrieval chunks.

02 · Privacy

A privacy layer in front of every model

Fourteen categories of personal and health data, plus each organization’s proprietary terms, masked before any query reaches a model; per-client vaults so nothing is shared between organizations.

03 · Trust

Grounding and verification

Reasoning layers that ground every answer in retrieved sources, verify cited references live, and check every calculation deterministically.

04 · Delivery

Production that stays up

Channel-native deployment with automatic retries and exponential back-off, so organizations reach their model where their people already work — around the clock.

Selected Work

Built fast,
and built to be checked.

Before and alongside Neusora: systems that hold AI to hard evidence — telemetry, sensors and market data.

Lead architect2025–26

LogScope AI

Intelligent diagnostics for security logs

An AI platform for large-scale cybersecurity log analysis. A verification engine audits the agent’s conclusions against deterministic system telemetry before anything is shown, and a deterministic-to-neural pipeline parses very large log files entirely in the browser, so the raw logs never leave the machine.

Anomaly Galaxy — 3D clustering of outliers
Forensic Hunter — vulnerability and attack detection
Chrono-Heatmap — spikes over time

Hungarian Embassy R&DNov 2024

Edge-AI plant diagnostics

Built in an 11-day sprint

A proof of concept pairing edge hardware with computer vision for round-the-clock, autonomous monitoring of plant disease, with causal machine-learning logic that maps a disease’s cause and forecasts recovery.

Co-presented to the Hungarian Ambassador with the principal investigator
Supported the first phase of a bilateral funding grant

1st PrizeFeb 2024

Currency-exchange forecasting

Al Mulla Exchange AI Hackathon

A time-series model forecasting currency exchange rates, built alone in a four-day hackathon — data ingestion, model design, interface and the final pitch.

First place as a solo entrant
Against 20+ teams of four

Expertise

From research
to production.

The range his work covers — from designing experiments and models to running them for organizations that depend on them every day.

Language-model systems

Retrieval-augmented generation
Embeddings and vector search
Fine-tuning (SFT, PEFT / LoRA)
Multilingual NLP, English and Arabic

Trust and privacy

Personal-data redaction
Per-client data isolation
Deterministic verification
Evaluation and benchmarking

Deep learning

CNNs, LSTMs and attention
Squeeze-and-excitation blocks
Multimodal sensor fusion
Anomaly detection

Computer vision and edge AI

Image-based diagnostics
Edge-hardware prototyping
Causal machine-learning logic
Time-series forecasting

Production engineering

Multimodal ingestion and OCR
Channel-native deployment
Resilient, self-retrying services
Cloud, container and edge deployment

Research practice

Experimental design and simulation
Comparative model benchmarking
Peer review as corresponding author
Scientific writing

Education & Recognition

The record.

Degree, honors and the programs behind the work, in order of weight.

Education

Bachelor of Engineering
in Computer Engineering

ABET-accredited program · 240 ECTS equivalent · 2026

A grades in Artificial Intelligence, Machine Learning and the senior design project. He began leading published AI research in his second year, aged 19.

LanguagesEnglishArabic

2025
Competitive open-access fundingKuwait Foundation for the Advancement of Sciences · awarded on the merit of the research
2025
Published author and oral presenterS4IoT’25 symposium · IEEE and Springer Nature
2024
1st Prize, solo entryAl Mulla Exchange AI Hackathon · against 20+ teams of four
2024
Presentation to the Hungarian AmbassadorEdge-AI prototype · Hungarian Embassy R&D collaboration
2024
Selected for a fully funded exchange cohortChinese language and culture · Dalian University of Foreign Languages, China
2023
GHABGATHON innovation competitionCompetitor · in partnership with the Kuwait Foundation for the Advancement of Sciences
2025–26
Head Start and Steer Your CareerAdvanced training programs · INJAZ Kuwait
2022
Android applications bootcampSix-week intensive · Kuwait Codes (CODED)

Contact

Talk to
Yousef directly.

For research collaboration, partnerships, speaking or a Neusora engagement.

yousef@neusora.com→ LinkedIn ↗ Google Scholar ↗ ResearchGate ↗