RustWASMEmulator
Intel 8080 Emulator
Cycle-accurate Intel 8080 CPU emulator running Space Invaders in the browser via WASM, with interrupt handling and a live debug panel showing CPU state, disassembly, and memory.
Hi, my name is
Mustafa Zaki Assagaf. MLOps, Platform, and System Engineer.
I keep GPU clusters, models, and infrastructure fast and reliable for a living
About Me
My name is Mustafa and I love building things end to end. These days my focus is MLOps and platform engineering: GPU clusters, Kubernetes, and the plumbing that keeps models training and serving reliably. I'm also able to do backend and frontend engineering, work with cloud on GCP and AWS, and play with system engineering projects.
Right now I'm a platform engineer at AI&, keeping on-prem GPU servers reliable with end-to-end telemetry, Kubernetes, and infrastructure as code. You can also checkout my latest project, the Intel 8080 emulator. It runs Space Invaders, and you can try it right in your browser!
Stuff that I have built
RustDockerPaaS
Pacil Web Service
Platform-as-a-service built from scratch in Rust for Fasilkom UI. A single binary running the git server, reverse proxy, web dashboard, and Docker infrastructure behind 400+ student deployments.
CUDAC++GPU
CUDA 100 Days
100 days of CUDA kernel programming, one kernel a day: vector ops, image processing, and PMPP book exercises, all logged day by day on real GPU hardware.
Other interesting projects
Click a tag to filter, click it again to clear it.
Neoman
Postman-like terminal application to send and receive HTTP requests and responses, complete with keyboard navigation
Zeta OS
Hobby OS from scratch in Rust for x86_64, booted with Limine and built reproducibly with Nix flakes
Rust Redis Clone
Redis-compatible server with an async TCP event loop and RESP protocol parsing, for the CodeCrafters challenge
Raft Consensus
Distributed Raft consensus algorithm implemented in Rust, complete with log replication
DOS Raycasting Engine
2D raycasting engine in C for MS-DOS, rendering a first-person 3D view straight into VGA memory
Lisp Interpreter
Lisp dialect interpreter written from scratch in C with a full REPL, grammar built on a parser combinator
OCaml C Compiler
C compiler front-end in OCaml with a lexer, parser, and AST, built test-driven for a functional programming course
NotionAPI
Contributed to the most popular Notion API SDK for Go, a library with 575+ stars
URL Shortener
Simple URL shortener with a web UI. Shortening any URL and getting the original back, with Redis caching in front of Postgres
Anonymous Chat
Chat application that lets you talk anonymously in rooms you create, streaming messages through Kafka over WebSocket
Hompimpa
Real-time multiplayer version of the Indonesian traditional game hompimpa, scaling WebSocket connections with Redis pub/sub
SSO Proxy
Proxy server to sign in with Universitas Indonesia SSO credentials via a REST API call, built with a colly scraper
KS-YAML
Kambing Style YAML, a formatter and linter for writing human-readable, developer-friendly YAML
Leetbot
Discord bot to get daily and random LeetCode problems, still alive and kicking on a server near you
Moodletui
Moodle terminal user interface to navigate courses and read forums without leaving the terminal
Moodlefeed
Discord and Line bot that watches Moodle and sends notification updates when courses change
Notion Notes
Discord and Line bot to save notes straight into Notion through its API
Extract Kurikulum
Parse and convert university curriculum from PDF to JSON, because sometimes JSON is better than PDF
My experiences
List of all of my professional experiences, totalling around 5 years of experience.
Platform Engineer @AI&
- Own the reliability of the on-prem GPU server fleet, keeping training and inference workloads running around the clock
- Built end-to-end telemetry with dedicated GPU observability, from hardware metrics to application traces and alerting
- Provisioned and operate the Kubernetes cluster and CI/CD pipelines that power the ML platform
- Automated infrastructure as code with Pulumi and pyinfra for reproducible server provisioning
MLOps Engineer @GoPay
Largest e-wallet and financial ecosystem in Indonesia
- Led the training of Sahabat AI, Indonesia's first LLM, on a 64× Nvidia H100 Kubernetes cluster using vLLM, SGLang, and NVIDIA TensorRT
- Architected and deployed an inference platform running on 40× Nvidia H100 GPUs, serving thousands of concurrent requests with low latency
- Built LLM-powered dispute resolution agent processing 10K+ daily tickets, reducing support costs by 40% and boosting customer satisfaction by 52%
- Built automated TTS & STT loan collection system handling 10K+ daily voice calls, saving IDR 3B annually ($200K+)
Platform Engineer @University of Indonesia
- Built an in-house PaaS from scratch in Rust as a single-binary application, capable of 400+ concurrent deployments
- Replaced Heroku with a highly optimized self-hosted solution, reducing infrastructure costs by $2000 per month
- Optimized server usage with cold-start memory swap from RAM to storage, reducing idle memory usage by 30-50%
MLOps Engineer @Soca AI
AI stealth startup focusing on edge devices
- Developed a high-performance server using Rust and Axum for real-time image processing
- Implemented image preprocessing shaders using wgpu in Rust and WebGPU for GPU-accelerated computations
- Designed and optimized an inference server and client using PaddleOCR with ONNX for efficient text recognition
HPC Cluster Admin (SLURM) @University of Indonesia
Faculty of Computer Science
- Managed and optimized SLURM-based HPC environment on an NVIDIA DGX Station for parallel programming coursework and research
- Configured job scheduling, GPU resource allocation, and monitoring for multi-user workloads
Software Engineer @Rvere
Shopify store speed and performance optimization
- Drove 17% increase in sales by optimizing critical P95 PostgreSQL queries, reducing user abandonment
- Cut infrastructure costs by 80% by migrating Google Lighthouse audit process to serverless AWS Lambda
- Prevented multi-hour service disruptions through real-time Grafana alerting system implementation
Backend Engineer @SayaKaya
Mobile mutual funds application
- Created B2B enabler service for closing a multi-hundred billion Rupiah (7 figures USD) deal
- Built a backend service from scratch using Golang with the Fiber framework and OpenAPI documentation
- Designed a CI/CD pipeline using Google Cloud Build, deployed on Google Cloud Run
DevOps Engineer @Fotoyu
Photo marketplace platform
- Improved API response time by 46% by implementing Redis caching for PostgreSQL database queries
- Developed and automated CI/CD pipelines using AWS CodePipeline, AWS CodeBuild, and AWS EKS with Terraform
Let's Talk
If you want to reach out or just want to say hi, the easiest way is to email me directly. I read everything, even the weird ones.