Laith M. alkhalil-AlHamed
Research portfolio

Research, built to grow.

A home for current projects, working papers, and future research directions across computer science. The portfolio is currently anchored by kinship verification and reproducible machine-learning systems.

Current projectPublicationsDiscuss research
Research approach

Infrastructure before scale

I am interested in research that is technically rigorous and practical to build on. That means treating the dataset pipeline, experiment configuration, evaluation, and reporting as part of the research—not as disposable support code.

Make experiments reproducible
Compare methods on consistent foundations
Treat datasets and evaluation as first-class engineering work
Build tooling that supports follow-on research
At a glance
4
method families
220
inferred family groups
805
people in local dataset
2,318
images in local dataset

Current work focuses on kinship verification from face imagery, while the broader portfolio remains open to research collaborations across software engineering, systems, and applied AI.

01 · Projects

Active research

Two active directions · room for more
Active · 2026

Kinship Verification Toolkit

Research infrastructure for reproducible computer-vision experiments

This project turns a fragmented kinship-verification literature into a maintainable Python platform for running, comparing, and extending experiments across multiple algorithmic families.

Computer VisionBiometricsRepresentation LearningMetric LearningReproducible ML SystemsDataset Engineering
View repositoryRelated paper
Research question

Kinship verification is challenging because resemblance signals are subtle, age-dependent, and easily confounded by pose, lighting, and expression. The project focuses on building research infrastructure that makes these experiments cleaner, more comparable, and easier to extend.

What it contributes
  • • Designed a unified CLI, config system, and reporting layout for consistent experimentation.
  • • Integrated public benchmark workflows with support for a local private dataset adapter.
  • • Created a cleaner base for benchmarking, ablations, publication-ready reporting, and future model extensions.
Research in progress

Accessible Maps: Deterministic Geospatial Tool Use for LLM Map QA

Current research collaboration · Georgia Tech Sonification Lab

A research prototype for accessible map question answering. An LLM translates natural-language questions into structured plans, while deterministic GeoPandas and Shapely tools perform the spatial reasoning and produce auditable traces.

Geospatial ComputingMCPGeoPandas + ShapelyLLM EvaluationAccessible Map QA
90.3%
Evaluation 1
95.3%
Evaluation 3
3,154
MapQA questions
View repository
Research question

Can language models handle question decomposition while deterministic geospatial tools handle exact topology, distance, direction, and thematic reasoning?

Current milestone

The current evaluation suite studies generalization across map types, question families, planner models, and external benchmarks.

02 · Publications

Papers and manuscripts

Working papers, publications, and other research outputs will collect here as the portfolio develops.

Working paper

Kinship Verification from Face Imagery

2026 · Related project: Kinship Verification Toolkit

A draft manuscript centered on a unified research toolkit for kinship verification, bringing classical descriptors, metric-learning pipelines, deep models, and gated pair-representation learning into one reproducible experimental framework.

Unifies four kinship-verification method families behind one reproducible CLI and config system.Supports public benchmarks alongside a curated in-house dataset spanning age variation, family archives, and identical-twin subsets.Produces experiment outputs designed for paper-ready comparison, ablation studies, and follow-on model development.
Open PDFView code
03 · Research interests

Where I want to keep exploring

Machine learning & vision

Representation learning, biometrics, computer vision, and evaluation of models on difficult real-world data.

Computer VisionRepresentation LearningBiometrics

Research systems

Reproducible experimentation, dataset engineering, benchmarking infrastructure, and tools that make research easier to extend.

ReproducibilityBenchmarkingDataset Engineering

Software & AI platforms

Software engineering, data-intensive platforms, applied AI, and dependable systems that connect research to usable technology.

Software EngineeringAI SystemsData Platforms
PhD outreach & collaboration

Interested in the work or a related direction?

I am open to PhD opportunities and research collaborations across software engineering, systems, data-intensive platforms, applied AI, and computer vision.

Start a conversation