Adham Aboulkheir
PhD Researcher · Explainable AI · AI Engineer
adhamaboulkheir@outlook.com
adhamaboulkheir.io
linkedin.com/in/adhamaboulkheir
github.com/Adham5172001
Colchester, United Kingdom
Profile
PhD researcher in artificial intelligence at the University of Essex, working on explainable fuzzy classifiers for living neural biocomputers. Three years of industry experience delivering production machine learning at BT Group and Saudi Motorsport, spanning generative models, retrieval-augmented LLM systems, computer vision and real-time streaming inference. Published at IEEE WCCI FUZZ-IEEE 2026. Particular interest in problems where a model's decision has to be defensible, not merely accurate.
Education
PhD in Artificial Intelligence
2025 — PresentUniversity of Essex, United Kingdom
- Supervised by Prof. Hani Hagras and Dr. Michael Barros
- Research: explainable AI and fuzzy rule-based classifiers for living neural biocomputers grown on Multi-Electrode Arrays
Industrial MSc in Artificial Intelligence
2023 — 2024University of Essex, United Kingdom
BA in Computer & Communication Engineering
2019 — 2024Alexandria University, Egypt
Publications
A Fuzzy-Based Approach for Interpretable Spike Detection in Living Neural Biocomputers
2026IEEE WCCI · FUZZ-IEEE 2026 · Maastricht, Netherlands
- Interpretable fuzzy rule-based system for spike detection on Multi-Electrode Arrays
- 97.74% F1-score across six biologically diverse chips using 337 human-readable IF-THEN rules
- ANNIGMA feature selection with genetic-algorithm-optimised rule weights
Experience
AI Product & Backend Delivery Engineer
Jun 2025 — PresentThresholdXpert AI Coach
- Designed and built the athlete performance analysis backend, moving the product from prototype to a stakeholder-ready Phase 3 release
- Fatigue modelling using ATL/CTL training-load metrics and a personalised recommendation engine
- FastAPI endpoints for session upload and readiness scoring
AI Researcher
Oct 2023 — Jan 2025BT Group
- Built a DCGAN + Beta-VAE + Stable Diffusion pipeline expanding 50 real telecom images into 350,000+ synthetic training images, improving YOLOv8 mAP@0.5 by 9.6 percentage points to 0.943 across ten fault classes
- Evaluated generation quality with FID and Inception Score, and validated utility through a downstream mAP ablation study
- Built an LLM + RAG fault diagnosis system (LangChain, FAISS, FastAPI) achieving 87.3% resolution accuracy and enabling remote resolution without engineer site visits
Associate AI Software Engineer
Nov 2023 — Oct 2024Saudi Motorsport Company
- Built a real-time ML system processing 13 telemetry channels per lap — throttle, brake, tyre temperatures and G-forces
- Streaming anomaly detection, lap-time prediction and tyre degradation modelling over a Kafka transport layer
- Containerised with Docker and orchestrated with Kubernetes for race-weekend reliability
Selected projects
Biocomputer Explainability Framework
2026PhD research
- Electrode importance (ANNIGMA and permutation), fuzzy rule extraction, counterfactual explanations, Moran's I spatial autocorrelation and functional hub detection
Agentic AI with LangChain & LangGraph
2025Open source
- Multi-agent workflow system for autonomous multi-step execution with tool use, memory and human-in-the-loop checkpoints
Generative AI & LLM Pipeline
2025Open source
- LoRA/QLoRA fine-tuning for Mistral-7B, hybrid TF-IDF RAG, chain-of-thought prompting, faithfulness scoring and BLEU/ROUGE benchmarking — 78% reduction in hallucination against an ungrounded baseline
Predictive Maintenance Pipeline
2024Open source
- Rolling feature engineering with an XGBoost failure predictor at AUC-ROC 0.97 and 48–72 hour lead time, MLflow-ready
Interval Type-2 Fuzzy Classifier
2025PhD research
- IT2 fuzzy classification with GA-optimised interval-valued rule weights for modelling second-order uncertainty
Full index of 21 public repositories at github.com/Adham5172001
Research
Explainable AIFuzzy rule systems (T1 & IT2)Genetic algorithmsNeural biocomputingMEA analysisFeature selectionSpatial statisticsStatistical methods
Machine learning
Deep learningGenerative AILLMs & RAGComputer visionNLPTime seriesReinforcement learning
Engineering
PythonPyTorchTensorFlowScikit-learnHuggingFaceLangChainLangGraphFastAPIDockerKubernetesKafkaAWSMLflowGitLinux
Download as PDF Opens your browser's print dialog — choose “Save as PDF”. The print stylesheet reformats this page for A4, so the PDF always matches what is on the site.