Skip to content
View josephfayyaz's full-sized avatar

Block or report josephfayyaz

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
josephfayyaz/README.md

Yousef Fayyaz

Applied Machine Learning Engineer
NLP, multimodal computer vision, reinforcement learning, and MLOps

I build research-driven machine learning systems with an emphasis on reproducibility, evaluation quality, and repository clarity.

LinkedIn Repositories GitHub

Location Open to Work Focus

Profile

I focus on building ML projects that are technically solid and easy to evaluate: clear problem definition, structured training pipelines, measurable results, and documentation that helps other engineers review the work quickly.

My recent work spans:

  • sarcasm and sentiment classification across English varieties using encoder, decoder, and adapter-based NLP pipelines
  • multimodal wildfire burned-area prediction from pre-fire satellite, terrain, weather, and infrastructure inputs
  • robust locomotion transfer in MuJoCo with PPO, curriculum domain randomization, and entropy scheduling
  • local MLOps infrastructure using Kubernetes, Istio, Katib, and browser-based development workflows

Selected Work

Unified experimentation repo for sarcasm and sentiment classification across multiple model families and English varieties.

Stack: Python, Transformers, RoBERTa, DistilBERT, Mistral, QLoRA, VAAT

Highlights: config-first pipelines, custom heads, adapter tuning, evaluation, plotting, and error analysis on a 17,760 train / 2,428 validation dataset.

Multimodal geospatial computer vision project for predicting final burned area from pre-fire observations.

Stack: Python, PyTorch, remote sensing, EfficientNet, FPN

Results: best multimodal model reached IoU 0.4026 and F1 0.5741, versus a Sentinel-only baseline at IoU 0.1515 and F1 0.2632.

MuJoCo Hopper transfer-learning research on robust locomotion under dynamics shift.

Stack: Python, MuJoCo, PPO, robust RL, evaluation tooling

Results: PPO + CDR + ES improved cumulative return by 72% over vanilla PPO and achieved more than 4x the return of PPO with uniform domain randomization.

Laptop MLOps lab that provisions a private-cloud style ML workflow on top of a local Kubernetes cluster.

Stack: Python, Kubernetes, kind, MetalLB, Istio, Katib, Docker

Highlights: end-to-end local platform setup with deployment runbooks, architecture docs, browser IDE access, and experiment infrastructure.

What I Bring

  • end-to-end ML experimentation: data handling, training, evaluation, diagnostics, and reporting
  • research translation: converting academic ideas into reproducible implementations
  • repository quality: readable project structure, documentation, and evidence-backed results
  • technical range: NLP, computer vision, reinforcement learning, geospatial ML, and ML systems

Technical Toolkit

Area Tools
ML & AI Python, PyTorch, Transformers, QLoRA, Scikit-learn, Jupyter
Domains NLP, Computer Vision, Remote Sensing, Multimodal Learning, Reinforcement Learning
Systems Kubernetes, kind, Istio, Katib, Docker, Git
Frontend Vue 3, Pinia, Tailwind CSS

Roles I Am Targeting

  • AI/ML internship
  • Junior machine learning engineer
  • Applied machine learning engineer
  • Research engineering or research assistant role

Where To Start

If you are reviewing my profile for hiring, these are the best entry points:

  • sarcasm-detection-nlp for NLP experimentation and model comparison
  • WildFire for multimodal computer vision and geospatial ML
  • RL for reinforcement learning, robustness, and transfer learning
  • ML_Ops for infrastructure thinking and MLOps workflow design

GitHub Snapshot

GitHub stats Top languages


I am most interested in teams where strong implementation, clear documentation, and measurable ML results matter as much as model choice.

Pinned Loading

  1. RL RL Public

    Python 6 1

  2. WildFire WildFire Public

    Python 3

  3. sarcasm-detection-nlp sarcasm-detection-nlp Public

    Figurative Language AI Recognition

    Python 3 1