Manal Alsuwat
العربية

AI Engineer

Manal Alsuwat

Hands-on experience developing LLM-powered solutions using RAG, LLM APIs, and multi-agent architectures. Skilled in building, evaluating, and deploying AI systems with Python and modern AI frameworks, and in turning research concepts into practical applications.

Location
Saudi Arabia
Education
M.Sc. Computer Science — AI
Languages
Arabic (native) · English (fluent)

Download CV GitHub Google Scholar LinkedIn Email

01

Experience

  1. Jul – Aug 2026

    King Khalid University

    AI Excellence Fellowship — AI Researcher & Engineer

    • Conducted applied research on LLM-driven automation for cybersecurity threat hunting.
    • Designed and evaluated multi-agent LLM architectures on recorded attack datasets.
  2. Sep – Dec 2024

    Taif University

    Adjunct Lecturer — Computer Science Department

    • Taught AI and computer science topics and guided students in implementing machine learning projects.
  3. Jun 2021 – Feb 2023

    Imam Abdulrahman Bin Faisal University

    AI Research Assistant

    • Trained and evaluated deep learning models for deepfake detection.
    • Prepared datasets and ran model experiments to support research objectives.
  4. Sep – Oct 2022

    IEEE Access

    Research Reviewer

    • Peer-reviewed submitted manuscripts for technical accuracy, novelty, and scientific rigor.
  5. Jul – Aug 2020

    Data Science Special Interest Group

    Data Collector & Programmer

    • Built preprocessing and clustering scripts for Arabic sign-language speech datasets.
02

Projects

  • A local layer in front of any LLMContext sources in → only what the request needs, redacted and attack-free, outContextRRouter3 cheap checks:size · source · PII01Selectrelevant contentembeddings02Detectinjected attackssmall classifier03HidePII redactionPresidio + regex04Securetool-call guardallow-listLLM−60% input tokens at equal accuracy · 0 PII values leaked (22-task benchmark)Runs locally on CPU — no LLM call, zero cloud overhead

    ContextShield AI

    2026

    Context-Preprocessing Layer for Cheaper, Safer LLM Calls

    Built a local, zero-LLM layer that sits in front of any language model: an adaptive router keeps only the context a request needs, redacts sensitive data and isolates planted prompt-injection attacks — cutting input tokens by about 60% at equal accuracy on a 22-task benchmark.

    Live demo Source on GitHub
  • Five specialist agents, one supervisorWindows event logs in → analyst-ready threat-hunting report outEvent logs01DetectionHayabusa · Sigmarule hits02InvestigationMSTICPycorrelation03Threat intelMITRE ATT&CKOTX context04ExplanationLLM narrativeevidence-checked05Responseprioritisednext stepsReportEvery LLM claim is verified against cited evidence before it reaches the reportEvaluation: DeepEval faithfulness 0.91–0.92 on two recorded Mimikatz scenarios

    ThreatHuntAI

    2026

    Explainable Multi-Agent LLM Framework for Threat Hunting

    Designed a five-agent LLM pipeline for analyzing Windows event logs and generating evidence-grounded incident reports.

    Source on GitHub
  • Every message is judged before it is answeredLLM-as-a-Judge security gate → RAG pipeline → grounded answerMessageLLM-as-a-Judgescores risk 1–5SAFERetrieveChromaDB · MMRGeneratesales agent LLMGrounded answerTHREATClassify attackmulti-labelRefuse + logAirtable · CSVSession endedLangChain · ChromaDB · Hugging Face embeddings · DeepEval evaluation · Streamlit UI + CLI

    Secure RAG Sales Agent

    2026

    Retrieval-Augmented LLM Assistant

    Built a RAG pipeline (LangChain, ChromaDB, Hugging Face embeddings) with an LLM-as-a-Judge security layer detecting prompt injection and jailbreak attempts.

    Live demo Source on GitHub
  • Learning from unlabeled audio and videoSelf-supervised pretraining → robust audio-visual representationsAudiospectrogramsVideoframe sequencesSelf-supervised pretrainingspatial–temporal objectivesJoint representationaudio-visual embeddingSound classification ↑ accuracyPublished: Expert Systems with Applications (2025) · Neurocomputing survey (2025)

    Master’s Thesis

    2024

    Multimodal Data Representation using Self-Supervised Learning

    Implemented self-supervised learning techniques to generate robust audio-visual representations, improving accuracy in sound classification tasks.

    Published research — see Publications ↓
03

Publications

Nine publications, 2020–2025. Full record on Google Scholar.

  1. 2025
  2. 2025
  3. 2025

    Comparative Analysis of Machine Learning Algorithms for Predicting Drug Mechanism of Action

    International Journal of Computer Science & Network Security (IJCSNS), vol. 25(5)

  4. 2023
  5. 2022

    Prediction of Diabetic Retinopathy Using Convolutional Neural Networks

    International Journal of Advanced Computer Science and Applications (IJACSA), vol. 13(7)

  6. 2022

    Makkah Landmark Recognition Using Deep Learning

    2nd International Conference on Computing and Information Technology (ICCIT), IEEE

  7. 2021

    Detection of Cardiovascular Disease Using Machine Learning Classification Models

    International Journal of Engineering Research & Technology (IJERT), vol. 10(7)

  8. 2021

    A Survey of the Application of Artificial Intelligence on COVID-19 Diagnosis and Prediction

    Engineering, Technology & Applied Science Research (ETASR), vol. 11(6)

  9. 2020

    A Hybrid Model Based on Radial Basis Function Neural Network for Intrusion Detection

    International Journal of Advanced Computer Science and Applications (IJACSA), vol. 11(8)

Conferences

  • 2nd International Conference on Computing & Information Technology (ICCIT)Jan 2022University of Tabuk
  • 16th Learning & Technology ConferenceMar 2019Effat University
04

Education & Training

  • Umm Al-Qura University

    Master’s Degree in Computer Science — Artificial Intelligence · 2024

    GPA 3.97 / 4 · with Honors

  • Taif University

    Bachelor’s Degree in Computer Science · 2019

    GPA 3.72 / 4 · with Honors

Certifications

  • Microsoft Azure AI Fundamentals (AI-901)Jul 2026
  • IBM AI Engineering Professional CertificateDec 2025
  • IBM Generative AI Engineering Professional CertificateNov 2025
  • NVIDIA Certified Associate: Generative AI MultimodalJul 2025
  • NVIDIA Certified Associate: Generative AI and LLMsJul 2025
  • AWS Certified Cloud PractitionerJan 2025
  • Jr Data Scientist CredentialNov 2024INE

Specialized training

  • Agentic AI BootcampMay 2026Saudi Digital Academy (SDA)
  • Professional Training in Generative AIMay 2025Generative AI Academy (SDAIA & NVIDIA)
  • Samsung Innovation Campus — Artificial Intelligence CourseOct 2024Samsung · Misk Skills
  • Advanced Artificial Intelligence TrainingJan 2024KAUST
  • Introduction to Artificial Intelligence TrainingNov 2023KAUST
  • HCIA-AI V3.0 Course — Huawei Certified ICT Associate in Artificial IntelligenceApr 2023Huawei
  • Machine Learning Engineering TrainingApr 2020Saudi Digital Academy (SDA)
05

Awards

  • 1st Place — Speech Clustering for Arabic Sign Language2020Data Science Special Interest Group
  • 1st Place — Best Arabic Research Content2018Etalem Tech Group
06

Contact

Happy to talk about generative AI and LLM systems, AI agents, or research collaboration.