Skip to content
View mzdwedar's full-sized avatar

Block or report mzdwedar

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.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. 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
mzdwedar/README.md

Hi, I'm Mohamed 👋

MLOps Engineer with 3+ years building self-service ML platforms and production LLM serving across AWS and Azure, covering data validation, lineage, CI/CD, serving and monitoring.

What I work on

  • LLM inference & serving – vLLM, NVIDIA Triton, quantization, KV/prefix caching, benchmarking
  • ML platforms & MLOps – platform, pipelines for ML workloads
  • AI agents – durable, human-in-the-loop workflows

Previously at Revic AI (2022–2025)

  • Led the Azure ML → SageMaker move and built a self-service ML platform (SageMaker, ClearML, GitHub Actions)
  • LLM serving on SageMaker with an Unsloth runtime inside NVIDIA Triton
  • Distributed AI-agent infrastructure on FastAPI, ECS, SQS and DynamoDB

Stack

Python PyTorch vLLM NVIDIA Triton AWS SageMaker ECS Azure ML Docker Kubernetes Terraform FastAPI LangGraph Temporal ClearML MLflow Databricks / Spark GitHub Actions Prometheus Grafana Sentry

📫 LinkedIn · mzjdwedar@gmail.com

Pinned Loading

  1. LLM-Inference LLM-Inference Public

    vLLM on a single 24GB L4: 5.8× throughput, p95 TTFT 89s → 0.8s via chunked prefill, prefix caching and FP8. Llama-3.1-8B, RAG workload.

    Python