Malek Swilam

|

Computer Engineering student at George Mason University's Honors College with a passion for machine learning, computer vision, and robotics.

Get In Touch View Resume
Scroll for more about me

About

Hey, I'm Malek Swilam, and I'm studying Computer Engineering at George Mason University's Honors College.

I did ML research at GMU; co-built OCRadar, an AI app that screens for oral cancer with 98% accuracy; and led software development for an FTC robotics team.

Outside of tech, I'm an Eagle Scout, where I built leadership, community service, and a love for the outdoors.

Experience

Machine Learning Intern

Yanez Compliance ↗
Jun – Aug 2026 Remote

Face Passive Liveness & Anti-Spoofing

  • Built a face anti-spoofing ensemble combining a gradient-boosted model on 50+ engineered image cues with a CNN-based depth-predicting model, cutting error 75% versus the deployed rule-based system on unseen data
  • Designed a rank-based score fusion for models with incomparable output scales, adding 3.4 AUC points over the stronger component and catching 100% of real attacks that had bypassed the production system
  • Ran an ablation study that isolated a single feature family as the cause of a cross-dataset failure, correcting it in 8 of 8 model configurations
  • Developed a CNN-based depth model for face liveness detection, then distilled it into a smaller network that held the same accuracy
  • Built the evaluation harness reporting standard biometric metrics (APCER, BPCER, ACER, AUC) across in-domain and cross-domain test sets

Adversarial Document Tampering Pipeline

  • Automated document tampering to generate adversarial data at scale, stress-testing document and identity verification systems against forgery attacks
  • Increased pipeline throughput 500%+ through multi-threading and a load-aware scheduler routing workers to idle models, backed by a FIFO job queue for batch runs
  • Replaced general-purpose generative models with a dedicated image-to-image face-swap model that produced higher-fidelity, more consistent output at 85–93% lower cost per image
  • Eliminated 100% of API cost for non-Latin text tampering by replacing model-based generation with a fully local deterministic renderer, expanding writing-system coverage 5×
Computer Vision Deep Learning PyTorch Biometrics Anti-Spoofing Model Evaluation Dataset Engineering Gradient Boosting Distributed Processing Python

Engineering Intern

Alarm.com ↗
May 2026 Tysons Corner, VA · On-site
  • Performed hands-on hardware lab work including soldering buck converters and embedded boards, bringing up low-level power circuits and embedded components
  • Assembled RJ45 Ethernet cables and provisioned Raspberry Pi units, building out the physical test automation infrastructure used across the hardware QA team
  • Configured and integrated home security devices (locks, contact/tilt sensors, lighting, cameras, thermostats, doorbells, and Flex I/O) into automated testing workflows across hardware QA cycles
  • Built an internal tool for lab device management so the hardware QA team could track and provision devices in one place
Hardware QA Embedded Systems IoT Soldering Raspberry Pi AI Tooling Python

Machine Learning Research Intern

George Mason University, ASSIP ↗
Jun – Sep 2025 Fairfax, VA · Hybrid
  • Built a full ML pipeline (data preprocessing, training, evaluation) for short-term rainfall forecasting, implementing and comparing LSTM, XGBoost, and ensemble models evaluated via MAE, RMSE, and R²
  • Improved rainfall prediction accuracy by 12–15% over individual models by combining LSTM and XGBoost into an ensemble, achieving 96%+ of predictions within ±1 mm of actual
Python PyTorch LSTM XGBoost ML Pipelines
View Abstract ↗
2023 – 2026 · 2 Seasons Vienna, VA
  • Led all robot software development in Java across two varsity seasons; mentored junior programmers in Java, Git, and FTC SDK over summer offseason
  • Built autonomous localization pipeline fusing wheel odometry with IMU for pose estimation; integrated AprilTag detection for field-relative pose correction; implemented motion profiling and trajectory planning via Road Runner
  • Developed a multi-mode vision targeting system using a Limelight camera backed by a finite state machine; implemented a PD controller on angular error for closed-loop real-time turret aiming
  • Designed multi-subsystem PIDF control: turret with adaptive static friction and velocity feedforward; flywheel velocity controller with piecewise-linear distance-regression model for range compensation
  • Implemented a 3-DOF PD controller with field-to-body-frame error projection for pose stabilization during shooting; tuned all PID coefficients
  • Qualified for State Championship both seasons and World Championship one season (ranked 8th seed captain in division)
Java Computer Vision PIDF Control Sensor Fusion Localization & Path Planning Road Runner Git / GitHub
DECODE Season Code ↗

Projects

Desk Calendar Agent

You text it "lunch with John Appleseed on Friday at noon." A 3B model running on the Raspberry Pi creates the event through the Google Calendar API, and deterministic checks stop anything destructive before it lands.

Local LLM Prompt Engineering Quantization Edge Inference Agent Safety Tool Calling Prompt Caching Model Evaluation Speech Recognition Embedded Linux Python
View project
Sentry app icon

Sentry

In active development

It sits in the macOS menu bar and reports what the machine is actually doing, with iPhone and Apple Watch companions. An MCP layer lets coding agents check the Mac has capacity before they start something heavy.

Systems Programming Distributed Systems Hardware Control Privileged Daemons Time-Series Data LLM Tooling AI Agent Infrastructure MCP Swift
Coming soon View on GitHub

Voice-Controlled Robotics Agent

You speak a command in plain English, like "stack the red cube on the blue cylinder." Claude Sonnet breaks that into ordered steps and a Franka Panda arm works through them in a live physics simulation. When a step goes wrong it reads the new scene and replans.

Python MuJoCo Voice Control Agentic AI Embodied AI Anthropic API LLM Planning FastAPI
View project
OCRadar logo

OCRadar

A cross-platform app that screens for early signs of oral cancer from a phone camera, using an Azure AI Custom Vision model trained on thousands of labeled clinical samples. Built by a team. I joined in 2024 as COO and lead developer for Android and Windows.

Python Computer Vision Azure AI Custom Vision Swift SwiftUI Flutter Dart
View project
View All on GitHub

Skills

Languages
Java Python Swift Dart JavaScript HTML / CSS
Libraries & Frameworks
PyTorch XGBoost NumPy FastAPI
AI & Agents
Local LLM MCP Tool Calling Agent Safety LLM Planning Anthropic API Edge Inference
Tools
Git / GitHub Ollama Azure AI MuJoCo FTC SDK
Hardware
Soldering Raspberry Pi SPI / GPIO E-ink Displays Embedded Linux
Focus
Machine Learning Computer Vision Robotics Biometrics

Education

Current 2026 – 30

Current · Fall 2026

George Mason University ↗

Honors College

B.S. Computer Engineering

Expected May 2030 · Fairfax, VA

Relevant coursework: Calculus III · Object-Oriented Programming · Intro to Engineering · Intro to Electrical & Computer Engineering

Concurrent 2022 – 26

Oakton High School ↗

Advanced Studies Diploma

Aug 2022 – Jun 2026 · Vienna, VA

  • Advanced Data Structures & Algorithms
  • AP Computer Science A
  • AP Computer Science Principles
  • AP Physics C: Electricity & Magnetism
  • AP Physics C: Mechanics (exam only)
  • AP Physics 1
  • AP Calculus BC
  • AP Statistics
  • AP Pre-Calculus
  • STEM Advanced Engineering
  • Honors Chemistry
  • AP US History
  • AP US Government & Politics
  • AP World History
  • AP Macroeconomics
  • AP Microeconomics

Northern Virginia Community College ↗

Non-Degree Seeking · Dual Enrollment

GPA 4.0

Aug 2024 – 2026 · Annandale, VA

Dual Enrollment English 12 · Dual Enrollment Entrepreneurship

Contact

Open to opportunities, projects, or just reaching out.

Say Hello
LinkedIn GitHub Resume