Joseph Nawar APPLIED AI · MACHINE LEARNING
AI ENGINEERING / MACHINE LEARNINGCAIRO, EGYPT  /  REMOTE-FIRST

I turn models into reliable, evaluated systems — from Arabic RAG and document automation to multimodal ML. I’ve completed a Computer Science degree (final results pending) and now teach as a university Teaching Assistant.

SCROLL TO EXPLORE
03 FEATURED BUILDSRESEARCH AUTHOR · MIUCC 20262026 · UNIVERSITY OF LONDON
01 / RESEARCH & PUBLICATIONS

Research &
publications.

Research isn’t just a line on my résumé. Here is the question, the evidence, and the limits of what the results can support.

Peer-reviewed · Accepted & presentedMIUCC / 2026

RuralStroke-Triage: A Hybrid Multimodal Late-Fusion Framework for Smartphone-Based Prehospital Stroke Screening

Joseph Nawar & Ayman Atia
6th International Mobile, Intelligent, and Ubiquitous Computing Conference (MIUCC 2026)

The research explores how incomplete smartphone-compatible evidence can be combined through auditable late fusion. The prototype evaluates visual, speech, and metadata branches on independent public proxy datasets, plus predefined symptom and fusion logic.

04evidence sources
33/33symptom-rule checks passed
276/276fusion checks passed

These checks verify implemented rules, not clinical diagnostic accuracy. There is no end-to-end validation on paired patient data.

Manuscript PDF withheldResearch case study
ORIGINAL PROJECT ARTWORK
Original RuralStroke project artwork showing four independent evidence branches and human-reviewed late fusion
RESEARCH / 01

Publication status: Accepted and presented — IEEE Xplore publication forthcoming. No DOI is claimed.

02 / FEATURED ENGINEERING

Things I've built.

Real implementations, measured results, and honest engineering trade-offs. Pick a project to see the details.

01 / AI AUTOMATION & BACKENDFEATURED

OpsFlow AI

Reliable purchase-order automation—from document ingestion and structured LLM extraction to human approval and ERP/CRM synchronization.

Built for recovery, traceability & reviewIdempotent workflows · validation · audit trails
FASTAPIPOSTGRESQLN8NODOO
Explore case study
01. SYSTEM / AUTOMATIONIllustrated workflow showing purchase order extraction, human review, and external integrations
02 / RETRIEVAL & LANGUAGERAG / ARABIC NLP

Kawaneen.

Arabic legal AI with hybrid dense/BM25 retrieval, reranking, citation-grounded answers, and safe abstention.

0.684 nDCG@10vs. 0.639 baseline · frozen retrieval evaluation
QDRANTRAGFASTAPIDOCKER
Explore case study
02. RETRIEVAL / KNOWLEDGEIllustration of Arabic search results and a grounded citation-based answer
03 / MULTIMODAL RESEARCHML / HEALTHCARE

RuralStroke Triage

A non-diagnostic prehospital screening research prototype combining vision, speech, symptoms, and metadata with interpretable late fusion.

0.9809 ROC-AUCVisual branch · held-out proxy-image evaluation
PYTORCHMOBILENETV2LATE FUSION
Explore case study
03. EVIDENCE / MULTIMODAL MLIllustration of the independent visual, speech, symptoms, and metadata branches combining in late fusion

Original explanatory graphics; not screenshots of deployed products. Quantitative results are scoped to the evaluations described in the case studies.

03 / BACKGROUND

Experience &
foundations.

A foundation in machine learning, software engineering, and helping others understand complex technical ideas.

Experience

2026 — PRESENT

Teaching Assistant

European Universities in Egypt · University of London CS programmes

Prepare Python and p5.js practical exercises for Intelligent Signal Processing; help students understand audio-processing algorithms, debug implementations, and reason about results.

SUMMER 2025

Machine Learning Intern · Remote

Elevvo Pathways

Built sales-forecasting regression and loan-approval classification workflows, from cleaning and model training to task-appropriate evaluation.

SUMMER 2024

Software Development Intern

Summit Technology Solutions

Contributed to C#/.NET development and designed SQL-oriented data models for a mock school database.

Education

BSc Computer Science

Machine Learning and Artificial Intelligence · University of London

Teaching centre: European Universities in Egypt

Studies completed 2026. Final results pending; expected First Class Honours.

IGCSEs

The Continental School of Cairo · Multiple A* grades

Certifications

IBM AI Engineering Professional Certificate
Stanford Machine Learning Specialization · Andrew Ng

AI & model development

PyTorch, TensorFlow, Scikit-learn, multimodal machine learning, reproducible evaluation, Pandas, NumPy

LLM applications

Retrieval-augmented generation, BM25 + dense retrieval, reranking, grounded QA, structured extraction

Software & integrations

Python, SQL, TypeScript, FastAPI, PostgreSQL, Docker, Qdrant, n8n, Odoo, HubSpot, GitHub Actions

04 / LET'S CONNECT

Let’s build AI
that matters.

Hiring for applied AI or machine-learning engineering? I’m interested in remote-first teams, research collaborations, and carefully scoped engineering work. My inbox is open.

joe.hany2005@gmail.com

Elsewhere on the internet

Based in Cairo, Egypt.
Interested in remote-first opportunities.