Click anywhere to skip
Ilham Rizanto · Software & System Builder

CRAFT

Statistics taught me to distrust assumptions, and building taught me to ship anyway. I look for the real problem, strip away everything that doesn't serve it, and build the smallest system that makes the manual work disappear.

PYTHONDJANGOTYPESCRIPTSQLAI & RAGAUTOMATIONDATA ANALYSISDOCKER

MINDSET

Start from first principles, then subtract. Simplicity is a form of respect for whoever maintains the system next.

Ilham Rizanto

“Solve the problem. Let the tools follow.”

— Ilham Rizanto
Location

INDONESIA

North Halmahera · North Maluku

SELECTED WORK

Featured projects

Every project started from a real bottleneck I ran into—never built just for the sake of building. All are internal or personal, non-commercial systems. Select any card to inspect the interface and architecture.

Show More (12)

AUTONOMOUS AI ECOSYSTEM

Not an app. An ecosystem.

NOVA is where my work on autonomous agents converges into one living ecosystem: an AI-run organization (RizaCorp) powered by the Hermes runtime, a spatial topology map, a local-first control plane, and native clients across every device I own. It is private and closed-source—what you see here is the ecosystem architecture in action.

Private Ecosystem · Architecture Showcase · Video & Live Telemetry Ready
NOVA ECOSYSTEM TOPOLOGY · 5 INTERCONNECTED LAYERS
01THE LIVING MAP

Neural Topology

Before anything else, I needed to see the entire ecosystem at once. Neural Topology renders the living ecosystem as one spatial knowledge graph: the orchestrator at the center, 12 divisions on the inner ring, and 24 specialist agents as leaves.

Every node carries live telemetry: which agent is active, which model powers them, how they connect, and what they last executed.

Radial ecosystem graphLive agent & division statePer-agent model visibilityMulti-agent Orchestration
NOVA
L0 Orchestrator L1 · 12 Divisions
02THE AUTONOMOUS ORG

RizaCorp & Hermes

Hermes is the multi-agent runtime. RizaCorp is the autonomous organization built on top of it: 24 AI specialists structured across 12 divisions, coordinated by a Chief of Staff that delegates, escalates, and consolidates reports.

They live where I already communicate. Each specialist is addressable directly inside our corporate Telegram group, so delegating complex work feels like messaging a team, not operating a dashboard.

24 specialists · 12 divisionsHermes runtimeTelegram group collaborationLLM APIs
telegram-group.png9:19 Portrait
03THE CONTROL PLANE

NOVA Control Plane

A local-first control plane within the ecosystem that decides what should happen and which node should execute it. Everyday commands resolve offline in under a millisecond with zero tokens; open-ended tasks route to a multi-step reasoning engine backed by persistent memory.

Within this ecosystem, the LLM is strictly an interpreter: it proposes structured parameters, while the command registry holds sole authority over what may run.

PythonFastAPISub-1ms zero-token NLULetta persistent memory
  • 01
    Voice or text intentINPUT

    Colloquial Indonesian voice (id-ID) or text from any client node

  • 02
    Tri-modal task brokerBROKER

    Routes by complexity: instant reflex, multi-step ReAct, or division delegation

  • 03
    Zero-token NLU & reasoningNLU / ReAct

    Sub-1ms offline NLU first; Letta persistent memory + LLM when reasoning is needed

  • 04
    Command registry gateREGISTRY

    Only whitelisted, schema-valid parameters pass; zero arbitrary shell execution

  • 05
    Target device nodeACTUATOR

    Re-validates locally, executes action, and streams honest telemetry back

04THE DEVICE MESH

Native nodes on every device

A Windows workstation agent with a Dynamic-Island style top-notch HUD, paired with native Android nodes for phone and tablet. Every client maintains an outbound TLS WebSocket link to the control plane, so one spoken sentence on the phone can orchestrate the PC—or vice versa.

Voice interaction runs full-duplex in natural Indonesian (id-ID) with barge-in support, while the visual HUD stays invisible until summoned.

PySide6 (Windows HUD)Kotlin & ComposeFull-duplex voice (id-ID)Cross-device WSS mesh
PCWINDOWS WORKSTATIONPySide6 Top-Notch HUD · Win32PHONEANDROID PHONEKotlin · Compose · Full-Duplex VoiceTABANDROID TABLETSpatial Companion · Always-On WSSNOVAWSS MESH
05THE GUARDRAILS

High agency, strict boundaries

An ecosystem with real actuators across personal devices demands uncompromising boundaries. Nothing executes outside the whitelisted command registry: the control plane validates the schema first, and the target device validates it again before touching the OS.

Execution status is never hallucinated: every command transitions deterministically through accepted, dispatched, running, succeeded, or failed.

Whitelisted command registryServer + client dual validationZero raw shell from LLM outputDeterministic state machine
dual-layer registry · honest lifecycle
ZERO SHELL EVAL
› “Activate work mode on my PC.” (voice · id-ID)
[accepted] offline NLU matched intent in 0.4ms · 0 tokens
[validated] registry schema OK · target=windows-workstation
[dispatched] WSS JSON-RPC 2.0 frame -> windows-agent
[running] local re-validation passed · executing 3 actuators
[succeeded] 3 / 3 device state changes verified honestly

TOOLING

Automation tools

Not every solution needs to be a full application. Some are small tools that quietly remove repetitive work.

BROWSER AUTOMATION

Data Collection Automation

Repetitive data-collection chores, turned into a menu you can double-click.

  • Scrapers pull progress recaps and raw dashboard exports from web systems, then push them straight into Google Sheets.
  • Drives a real local Chrome session over the DevTools protocol, and recovers on its own when a session expires.
  • Applies bulk follow-up to flagged records, so nobody has to click through hundreds of them by hand.
  • Wrapped in a single interactive menu, so anyone on the team can run it without touching a terminal.
PYTHONPLAYWRIGHTCDPPANDASGOOGLE SHEETS API

WHAT I WORK WITH

Skills & tools

Languages, frameworks, and infrastructure built across real internal systems—picked for what each problem actually demanded.

LANGUAGES

PythonTypeScriptJavaScriptPHPSQLRKotlin (Android)
The stack follows the problem● Always learning

BACKEND & DATA

Django & DRFFastAPINode.js & Expresspandas & NumPyPostgreSQL & SQLiteRedisJWT & Role-based Access
The stack follows the problem● Always learning

FRONTEND & MOBILE

React & ViteNext.jsTailwind CSSHTMX & Alpine.jsCapacitorLeaflet & GeoJSONStreamlit
The stack follows the problem● Always learning

AI & AUTOMATION

RAG & EmbeddingspgvectorLLM APIsMulti-agent OrchestrationPlaywrightDocument AutomationGoogle Workspace APIsDocker & Self-hosting
The stack follows the problem● Always learning

HOW I THINK

Working principles

Not a list of commercial services—just how I approach problems, and what I value most when collaborating with others.

Macro photograph of precision camera lenses on a dark matte surface
PRINCIPLE 01

Problem first

A well-defined problem is already half solved. I get the problem clear before choosing anything, and the technology follows. If it requires something I haven't used yet, I learn it.

• First principles• Stack-agnostic
Minimalist compact hardware layout with zero redundant parts
PRINCIPLE 02

Simplicity & parsimony

The simplest system that does the job wins. Fewer moving parts mean fewer ways to break and far less burden to maintain.

• Less is more• Maintainable
Interconnected server racks and structured fiber-optic network cabling
PRINCIPLE 03

Integrated systems

Good tools talk to each other. I connect spreadsheets, databases, bots, and dashboards so the entire workflow fits together instead of living in isolated silos.

• Integration• End-to-end
Macro photograph of a high-density silicon processor and precision circuitry
PRINCIPLE 04

Leverage, not autopilot

AI lets one builder ship what used to take a team. I use it to move faster, then verify every result myself. Speed is the gain; judgment and accountability stay with me.

• AI as leverage• Verify the output

RESUME

Curriculum vitae

A concise overview of my background, selected systems, and technical stack in one document. The PDF edition is currently being finalized and will be available here soon.

Coming soon

BACKGROUND

My journey

From statistical training in Jakarta to building real-world systems in North Halmahera.

Photo Slot Ready

/public/images/journey/bps-work.jpg

Supports .jpg, .png, or .webp — automatically renders once placed

Statistician — BPS-Statistics Indonesia
Field & statistical operations · BPS-Statistics Indonesia

Statistician

BPS-Statistics Indonesia

North Halmahera, Indonesia

2024 — PRESENT
  • Work with official data, statistical analysis, and public-sector workflows every day—while solving the practical IT and engineering bottlenecks around them.
  • Build internal systems for field tracking, data cleaning, performance reporting, document automation, and AI-assisted search when manual workflows prove too slow or error-prone.
  • All software showcased in this portfolio represents personal, non-commercial initiatives—not official institutional products.

Photo Slot Ready

/public/images/journey/stis-graduation.jpg

Supports .jpg, .png, or .webp — automatically renders once placed

Bachelor of Applied Statistics (S.Tr.Stat.) — STIS Statistics Polytechnic · D-IV Statistics
Graduation · Politeknik Statistika STIS

Bachelor of Applied Statistics (S.Tr.Stat.)

STIS Statistics Polytechnic · D-IV Statistics

Jakarta, Indonesia

2020 — 2024
  • Trained in applied statistics, sampling methodology, econometrics, and quantitative problem-solving.
  • Started with Python, R, and SQL for statistical computing—then expanded into full-stack engineering and AI orchestration as real problems demanded broader tools.

LET'S CONNECT

Have a problem worth solving?

I'm not job hunting, but I'm always glad to meet builders, partners and people with interesting problems. Open to networking, collaboration, and the occasional side project when time allows.

Say Hello
Email: ilhamgb11@gmail.com
GitHubLinkedInInstagramXThreads