PhD Researcher · Computer Science

Fatima Farooq

Graduate Research Assistant, Large-Scale Intelligent Data Systems (LIDS) Lab

University of Texas at Arlington · Advisor: Prof. Ashraf Aboulnaga

Research Interests: Relational Deep Learning · Relational Foundation Models · Graph Neural Networks · Knowledge Graphs · Scalable Machine Learning

fxf3214 [at] mavs.uta.edu  |  fatimafarooq2310 [at] gmail.com

CV Email LinkedIn GitHub
Fatima Farooq

About

I am a PhD student and Graduate Research Assistant in the Large-Scale Intelligent Data Systems (LIDS) Lab at the University of Texas at Arlington, advised by Prof. Ashraf Aboulnaga. My research focuses on machine learning over relational databases and large-scale knowledge graphs, with particular interests in relational foundation models, graph neural networks, and scalable learning.

Before beginning my PhD, I spent 2.5 years as a Senior AI Developer building production machine-learning systems for real-time radar, geospatial analytics, anomaly detection, and aircraft-trajectory tracking. I also have several years of university teaching and freelance AI/ML experience. I am interested in research that is technically rigorous, scalable in practice, and useful beyond a single benchmark.

Current Research

FORTE

Relational Foundation Models

A relational foundation model designed to learn directly from multi-table databases and transfer across unseen databases and downstream tasks without task-specific retraining.

gHAWK

Scalable GNN Training on Knowledge Graphs

Structural encoding for scalable GNN training on large heterogeneous knowledge graphs, combining precomputed structural priors with lightweight graph learning.

Relational Data Selection

Efficient Training and Inference

Studying how to retrieve and select the most informative slice of a large relational database for each prediction query under practical memory and latency constraints.

Large-Scale Graph Learning

Web-scale Benchmarks

Benchmarking models on large graph datasets including MAG240M, OGB-MAG, OGB-WikiKG2, RelBench, and other relational learning benchmarks.

Publications & Manuscripts

FORTE: A Relational Foundation Model with Foreign-Key-Role Awareness Under Review
Fatima Farooq*, Humera Sabir*, Ashraf Aboulnaga · 2026

A relational foundation model that learns directly from multi-table databases and transfers to unseen databases and tasks without retraining.

gHAWK: Structural Encoding for Scalable Training of Graph Neural Networks on Knowledge Graphs Under Review
Humera Sabir*, Fatima Farooq*, Ashraf Aboulnaga · 2026

Structural encoding for scalable GNN training on large knowledge graphs. Evaluated on MAG240M, OGB-MAG, and OGB-WikiKG2. arXiv

* Equal contribution.

Experience

Graduate Research Assistant

LIDS Lab, University of Texas at Arlington · Arlington, TX
Aug 2024 – Present
  • Develop relational and graph learning models for multi-table databases and large heterogeneous knowledge graphs.
  • Co-developed gHAWK for scalable GNN training on graphs with hundreds of millions of nodes; integrated with PyTorch Geometric and DGL.
  • Developed FORTE for prediction directly over relational databases, supporting classification, regression, recommendation, and link prediction.
  • Benchmark models at web scale, including MAG240M (244M nodes), OGB-MAG, and OGB-WikiKG2.
  • Administer shared GPU infrastructure, including user access, CUDA/driver upgrades, environments, storage, and job scheduling.

Senior AI Developer

Center of AI and Computing (CENTAIC), NASTP · Islamabad, Pakistan
Jan 2022 – Jul 2024
  • Led production ML pipelines for aircraft trajectory classification from streaming radar data, sustaining approximately 800–1,000 tracks per second.
  • Built anomaly detection over continuous sensor streams and integrated inference with PyQGIS for real-time geospatial visualization.
  • Developed a modular decision-support system combining trajectory classification, anomaly alerts, and geospatial context.
  • Deployed Memgraph and Cypher-based relational analysis over noisy geospatial data.
  • Designed distributed IPC components, stress-testing infrastructure, and cross-platform Qt applications.

Graduate Student Researcher

AIM Lab, FAST-NUCES · Islamabad, Pakistan
Sep 2021 – Aug 2022
  • Built a transformer-based Urdu speech-recognition system by fine-tuning wav2vec 2.0.
  • Achieved approximately 16% word error rate and 8% character error rate.
  • Curated an Urdu audio/transcription corpus and investigated dialect identification using word2vec and transformer encoders.

Lab Instructor

FAST-NUCES · Islamabad, Pakistan
Jan 2019 – Jan 2022
  • Taught Data Analysis, Data Visualization, Databases, Object-Oriented Programming, and Digital Logic Design.
  • Mentored student ML, programming, and data-pipeline projects.

Software Engineer, Deep Learning

VR & Motion Simulation Lab, UET Taxila · Taxila, Pakistan
Apr 2018 – Jul 2019
  • Built an autonomous driving simulator using OpenCV and deep learning for low-cost perception-data generation.
  • Implemented MobileNet-SSD object detection and camera-calibration-based lane detection.

Education

Ph.D. in Computer Science

University of Texas at Arlington
Aug 2024 – Present

CGPA: 4.00/4.00 · Advisor: Prof. Ashraf Aboulnaga · Merit-based STEM scholarship

M.S. in Data Science

FAST National University of Computer and Emerging Sciences, Pakistan
Sep 2020 – Aug 2022

CGPA: 3.74/4.00 · Silver Medalist · Graduated 2nd in cohort

Thesis: Transformer-based Automatic Speech Recognition for Low-Resource Languages

B.S. in Software Engineering

University of Engineering and Technology, Taxila, Pakistan
Oct 2015 – Aug 2019

CGPA: 3.48/4.00

Thesis: Autonomous Driving Simulator using OpenCV and Deep Learning

Selected Projects

B+-Tree Index & Query Executor

Implemented B+-tree indexing, buffer pool management, WAL/redo logging, and a C++ query executor with nested-loop join, projection, and aggregation.

Recommendation System at Scale

Collaborative-filtering and content-based recommendation engine over 25M records, with a D3.js analytics dashboard.

DBLP Bibliometric Prediction

Neural prediction and t-SNE visualization of publication venue trends, reaching approximately 93% accuracy.

Freelance AI & ML Consulting

Completed 200+ projects across deep learning, computer vision, applied data science, visualization, and ML engineering for global clients.

Awards & Honors

Teaching & Mentoring

Technical Skills

Languages Python, C/C++, SQL, Java, R, MATLAB, Cypher
ML / Deep Learning PyTorch, PyTorch Geometric, DGL, Hugging Face Transformers, TensorFlow, Keras, scikit-learn
Research Areas Relational Deep Learning, Graph Neural Networks, Knowledge Graphs, Transformers, Self-Supervised Pretraining, Few-Shot and Zero-Shot Transfer, Anomaly Detection, Speech Recognition
Data & Systems Distributed training, Apache Spark, Hadoop, Hive, Pig, Memgraph, pthreads, IPC, CMake, gdb, valgrind, perf, Qt
Infrastructure Linux, CUDA, GPU server administration, Conda, Docker, job scheduling, storage management, cluster monitoring
Database Internals B+-tree and LSM indexing, buffer pools, WAL / redo logging, joins, projection, aggregation
Benchmarks OGB, MAG240M, OGB-MAG, OGB-WikiKG2, RelBench, FB15k-237