Marchel FahreziContact
[00] FIELD SHEET · THE PERSONBANDUNG · [6.9175° S, 107.6191° E]

I design the autopilot.

I’m Marchel Fahrezi, “Lynx”. A developer in Bandung who leads a team, certified in cloud and data. Sometimes DevOps, sometimes IT, handle everything. RixLab is where I put the tools I build on my own time, and the work I take on.

← Back to RixLab

Read why

REF · HOME SERVER · FRONT

Drawn, not to scale. Red light means it’s on.

[01] WHY

I’ve always been the person things land on.

“Sometimes DevOps, sometimes IT, handle everything.” You can’t handle everything by hand, so I build systems that handle themselves. Then I design how they recover when they don’t.

I run my own home server (and clean up after it). I’d rather be at Ijen crater or in Baluran than in another meeting.

[02] Stack and numbers

Python · Tomo, Go · wok, Rust → WebAssembly · netlab, Runs Cisco IOS, RouterOS, FRR, Open vSwitch, x86, In a browser tab, Kubernetes (Rancher), Google Cloud, Microsoft Fabric

51×

wok vs a Playwright script, per command

5exams

Proctored: Microsoft, Google, Oracle

32labs

netlab, Beginner to Expert

7yrs

One company, from writing the code to leading the team

Live  Packets forwarded by the drawings on this page:0

[03] ORIGIN · SURVEY LINE

From a networking class to the autopilot.

I started in a networking class at SMKN 4 Malang in 2016. In 2019 I took my first job as a full-stack developer, and I’ve been the person things land on ever since: sometimes DevOps, sometimes IT, handle everything. I finished my degree at Widyatama while working full-time. In 2024 I started leading the team. You can’t handle everything by hand, so I build systems that handle themselves, and then I design how they recover when they don’t.

IJEN · WHERE I’D RATHER BE
Photo · portrait · TBD
  1. 2016

    SMKN 4 Malang

    Computer systems, networking, and telecoms. Where the packets started.

  2. 2019

    Full-stack developer

    Semesta Teknologi Informatika, Bandung. First job, August 2019.

  3. 2020–24

    B.Eng. Informatics, Widyatama

    A year into the job. Finished while working full-time.

  4. 2024

    Development team lead

    Since March 2024. At work I lead a team building AI-driven operations platforms for enterprise infrastructure.

  5. Now

    Agentic AI

    Agents that act, the tools they act with, and somewhere to run them.

[04] CERTIFIED · RANGE

Certified, and not only in code.

Five proctored exams across data and cloud, each one checkable at the issuer. Below them, every area I work in, and what proves it: an exam, a badge, or work you can read.

  1. Microsoft

    Fabric Data Engineer Associate

    Oct 2026 – Oct 2027

    Verify Microsoft Fabric Data Engineer Associate
  2. Microsoft

    Fabric Analytics Engineer Associate

    Sept 2026 – Sept 2027

    Verify Microsoft Fabric Analytics Engineer Associate
  3. Google

    Associate Cloud Engineer

    Feb 2026 – Feb 2029

    Verify Google Associate Cloud Engineer
  4. Oracle

    Data Platform 2025 Foundations Associate

    Oct 2025 · no expiry

    Verify Oracle Data Platform 2025 Foundations Associate
  5. Oracle

    Cloud Infrastructure 2025 Foundations Associate

    Oct 2025 · no expiry

    Verify Oracle Cloud Infrastructure 2025 Foundations Associate

Certified Badge + practice In practice

  • Certified

    Data engineering and analytics

    • Microsoft Fabric Data Engineer Associate
    • Microsoft Fabric Analytics Engineer Associate
    • Oracle Data Platform 2025 Foundations
    • Applied Data Science with Microsoft Fabric
  • Certified

    Cloud infrastructure

    • Google Associate Cloud Engineer, valid to 2029
    • Oracle Cloud Infrastructure 2025 Foundations
  • Badge + practice

    Networking and security

    • Build a Secure Google Cloud Network (skill badge)
    • Site-to-site BGP VPN to Azure
    • netlab: an L2/L3 fabric I wrote
  • Badge + practice

    AI and agents

    • OCI 2025 AI Foundations Associate
    • Building Gen AI Applications with Microsoft Azure
    • Tomo and wok, in the open
  • In practice

    DevOps and Kubernetes

    • High-availability RKE2 run from Rancher, Helm deploys
    • CI on three OSes for wok
    • “Sometimes DevOps” is in the job
  • In practice

    Systems administration

    • Active Directory replication across domain controllers
    • MSSQL Always On cluster on WSFC
    • A home server I run and clean up after
  • In practice

    Architecture and design

    • HA PostgreSQL with Patroni and etcd
    • Tomo: coordinator, remote workplaces, restart recovery
    • netlab: zero backend, everything in the tab
  • In practice

    Software development

    • Kofa, Arithmos, MBlast, SSPR: apps I led
    • Python, Go, Ruby, Rust → WebAssembly
  • Badge + practice

    Leading a team

    • Become an AI-Powered People Manager
    • Development team lead since 2024

[05] RESEARCH · PUBLISHED

Published with the error included.

Three papers, 2023 to 2024. The first one is a machine that watches Bandung’s traffic.

Field report · Bandung trafficJUTIF · 2024

Traffic Flow and Congestion Detection with YOLOv8 and ByteTrack-Based Multi-Object Tracking

Counting cars and spotting jams from the city’s existing traffic cameras.

Vehicles detected correctly~80%
Average error in the count~17%
Congestion called correctly~80%

Counts vehicles and flags congestion from Bandung's existing CCTV cameras. About 80% detection accuracy, about 17% average error in the count, and about 80% accuracy at calling congestion.

Jurnal Teknik Informatika (JUTIF), vol. 5 no. 4, August 2024. With Eka Angga Laksana.Paper · Code and weights

  1. Customer Segmentation and Analysis Based on Gaussian Mixture Model Algorithm

    Sorting customers into groups by how they spend, without deciding the groups in advance.

    Groups customer spending with a Gaussian mixture model. BIC, with a diagonal covariance, picks four clusters. They hold 32%, 30%, 27%, and 10% of customers.

    Widyatama International Conference on Engineering 2024. Atlantis Press, 29 December 2024, pp. 67–75. With Eka Angga Laksana.

    Paper →

  2. Consumers as Co-Creation for Determining Business Strategy Using the Apriori Algorithm in International-Scale Retail

    Finding which products people buy together, and using that to plan what a store does next.

    Mines international-retail baskets with Apriori, then checks the rules on later transactions. Five rules clear 80% confidence. The fit uses 75% of the transactions and holds out 25%.

    Jurnal Sistem dan Teknologi Informasi (JUSTIN), vol. 11 no. 3, July 2023. With Feri Sulianta, Yosi Malatta Madsu, and Yenie Syukriyah.

    Paper →

[06] HOW I THINK

Opinions I’ll defend.

Each one has someone who would argue. That is the point.

Disagree?Tell me why →

  1. Tomo shows me the plan before it splits the job. After “gas”, I walk away.

  2. Everything crashes. What matters is what it remembers. Tomo’s active tasks resume from the saved conversation and tool results.

  3. Not before it does one useful thing. wok is one binary. netlab is a browser tab.

  4. Publish the number and where it stops being true. wok is 51× faster per command. Inside one long Playwright script, it’s a tie.

  5. Few moving parts, obvious failure modes.

[07] NOW

What’s on the desk.

Buildingnetlab. This week it learned VLAN trunks, DHCP relay, OSPF between two routers, static NAT, and traceroute.

Stuck onRouterOS v7 wants a 64-bit CPU. The emulator netlab uses for x86 is 32-bit, so only RouterOS v6 boots. Getting past that means QEMU in WebAssembly.

— Lynx, 10 Nov 2026

  • netlabThe browser network lab.
  • skillsRules I give coding agents.
  • droppyA notch for Windows, in native Rust.
  • tomoThe agent swarm, in alpha.

[08] CONTACT

Want to talk?

Work, a tool, or a disagreement. Write to me.

  • Agents that finish the job
  • Tools with no setup guide
  • Systems that survive the restart
Email me

hi@rixlab.xyz
GitHub · Alg0rix · LinkedIn