Ziad Waleed

Machine Learning Engineer

https://talent.gravityer.com/ziad-waleed

Machine Learning Engineer with less than a year in AI Systems, TensorFlow & PyTorch.

Freelance

Key Strengths

Strong foundational knowledge in Deep Learning frameworks (PyTorch, TensorFlow/Keras) and Machine Learning algorithms (Scikit-Learn).
Demonstrated ability to build and deploy end-to-end ML systems, including API development (FastAPI), containerization (Docker), and cloud deployment (AWS EC2/S3, Hugging Face Spaces).
Extensive project experience in advanced ML topics such as deepfake detection (audio and video), large language models (GPT from scratch), and reinforcement learning (Super Mario Bros AI Agent).
Proficiency in data preprocessing, feature engineering, and handling complex datasets, including distributed training and memory optimization.
Experience with MLOps practices like CI/CD pipelines (GitHub Actions) and model artifact versioning (AWS S3).

Cultural & Operational Fit

Cultural Fit Analysis

The candidate's project diversity, including deepfake detection, LLM development, and reinforcement learning, shows a broad interest and adaptability within the ML domain. Their participation in competitive programming and freelance work indicates a proactive and self-driven individual. However, the limited professional experience (freelance and internship) and ongoing education suggest a need for mentorship and structured team environments, which might require careful integration into a senior role.

Soft Skills & Operational Fit

The candidate demonstrates strong problem-solving skills through competitive programming and project work. Their experience as a freelance copywriter suggests good communication and project management abilities. The focus on end-to-end system development indicates an operational mindset, capable of taking projects from conception to deployment.

AI is analyzing your overall score…

Identifying your key strengths…

Evaluating your skill match against the job requirements…

Assessing your cultural and operational fit

Projects

Generative Pre-trained Transformer (GPT) from Scratch

April 1, 2026 – May 1, 2026

Engineered a 10.7M parameter Decoder-only Transformer from scratch in PyTorch, implementing custom scaled dot product attention with causal masking, feed-forward networks, and decoupled weight decay AdamW for autoregressive text generation. Designed an End-to-End machine learning pipeline by decoupling the PyTorch inference engine and deploying the model via a FastAPI REST endpoint, serving a custom asynchronous web interface for real-time, dynamic prompt generation. Validated architectural effectiveness by developing a baseline Bigram language model, successfully proving the multi-head attention mechanism's ability to capture complex, long-range linguistic dependencies across a 256 token context window.

Deepfake Video Detection

March 1, 2026 – April 1, 2026

Architected a hybrid deep learning pipeline, developing a highly accurate 94% video classification model using EfficientNetB4 for spatial frame-level feature extraction and a stacked GRU network for temporal sequence modeling. Engineered a cloud-native REST API, building and deploying a FastAPI microservice on Hugging Face Spaces that optimizes memory allocation to seamlessly download, process, and analyze asynchronous video URL streams. Implemented enterprise-grade evaluation suites, addressing severe dataset class imbalances using weighted loss penalization to ensure real-world reliability against adversarial media.

Sonic Sleuth – Audio Deepfake Detection

January 1, 2026 – March 1, 2026

Engineered a custom PyTorch CNN based on state-of-the-art academic research, surpassing the original paper's baseline by achieving a 99.15% test accuracy and 98.98% F1-score on a diverse 55,000+ file dataset. Architected a distributed dual GPU training pipeline to handle complex LFCC and SpecAugment processing, leveraging larger effective batch sizes to achieve smoother gradient descent, optimal convergence, and model generalization. Deployed the inference engine to Hugging Face Spaces as a FastAPI microservice, safely streaming audio payloads directly into RAM while enforcing strict Pydantic JSON contracts and 10MB limits to prevent Out-of-Memory (OOM) crashes.

Super Mario Bros AI Agent

November 1, 2025 – December 1, 2025

Trained an autonomous Reinforcement Learning agent to play Super Mario Bros using Proximal Policy Optimization (PPO) via Stable Baselines3 and PyTorch. Engineered a custom OpenAI Gym environment pipeline, implementing frame stacking, grayscale transformation, and action space reduction to optimize the model's pixel-based state processing. Configured CUDA accelerated GPU environments in Python to drastically reduce training time, yielding an agent capable of independently navigating dynamic in-game obstacles and maximizing scores.

Key Strengths

Strong foundational knowledge in Deep Learning frameworks (PyTorch, TensorFlow/Keras) and Machine Learning algorithms (Scikit-Learn).
Demonstrated ability to build and deploy end-to-end ML systems, including API development (FastAPI), containerization (Docker), and cloud deployment (AWS EC2/S3, Hugging Face Spaces).
Extensive project experience in advanced ML topics such as deepfake detection (audio and video), large language models (GPT from scratch), and reinforcement learning (Super Mario Bros AI Agent).
Proficiency in data preprocessing, feature engineering, and handling complex datasets, including distributed training and memory optimization.
Experience with MLOps practices like CI/CD pipelines (GitHub Actions) and model artifact versioning (AWS S3).

Cultural & Operational Fit

Cultural Fit Analysis

Soft Skills & Operational Fit

Ziad Waleed

Key Strengths

Cultural & Operational Fit

About

Top Skills

Skills

Education

Experience

Projects

Key Strengths

Cultural & Operational Fit