Open to remote roles & freelance builds

I build AI products that show their work.

I’m Rifqi Sigwan Nugraha, a full-stack AI developer from Jambi, Indonesia, and co-founder of Plus The Site. I design the pipeline, write the backend, and wire up the tests, so my flagship builds come with a live demo, public source, and passing tests you can check yourself.

See the work Download CV

Jambi, Indonesia · UTC+7

verification log

test carecanvas
19 / 19 passing
test nalar
24 / 24 passing
cite research
IEEE Xplore · 1st author

67 passing tests across three public repositories
~300 households covered by an IoT flood early-warning system
10+ computer-vision models wired into one real-time dashboard
2× national finalist, Telkom Indonesia Innovillage

AI pipelines
Next.js
Backend architecture
Supabase
Agentic workflows
Python
System integration
PostgreSQL
Systematic QA
Flutter
IoT
TypeScript

Selected work

Two flagship builds. Open them up.

CareCanvas and NALAR each have a public URL and a public repository, so you never have to take my word for it. Run the demo, read the code, check the tests.

Every case study splits into what the product does, how it’s engineered, and exactly where its verified boundary sits.

Flagship · live public demo

01 / 02

Human-in-the-loop AI illustration pipeline

CareCanvas

Turns a visual brief into a child-appropriate illustration through bounded AI agents, an explicit human approval gate, and durable execution, with a full trace a reviewer can step through when something goes wrong.

Try the live demo Read the source

19 / 19: automated tests
2: bounded attempts
96: QA drill score
0: external provider calls

CareCanvas public demo showing its AI illustration workflow and review interface — Live interface · deterministic safe-demo mode

01
BriefStructured intent
02
SafetyStop before spend
03
ApproveHuman gate
04
GenerateLocal or Flux adapter
05
Visual QAOne bounded retry

What’s built

The hard boundaries live in code.

Explicit state transitions, Inngest approval waits, Claude and Flux adapters, owner-scoped Supabase RLS, raw-body Ed25519 webhook verification, and auditable traces.

What’s verified

The demo proves the orchestration.

The public demo runs on deterministic local adapters and makes zero external provider calls. The live Anthropic Claude, fal.ai Flux, Inngest, and Supabase paths stay environment-gated.

Next.js 16
Claude adapter
fal.ai Flux
Supabase
Inngest
TypeScript

Live prototype

02 / 02

Role-based decision support for cooperatives

NALAR

An operations prototype covering sales, POS, customers, receipt verification, national mapping, and an optional grounded AI assistant. Built so the core workflow keeps running even when the AI provider doesn’t.

Open the live app Read the source

24 / 24: tests passing
Strict: TypeScript
Graceful: degradation

Disclosure: NALAR runs on synthetic portfolio data. It is not presented as a production cooperative or as having live users.

Server-owned access

Supabase service-role calls stay behind server APIs. Elevated credentials never reach a public client.

AI as an option, not a dependency

Gemini answers are grounded in application context, with deterministic fallbacks when the provider is down.

Tested failure paths

Tests cover forecasting, tamper detection, fail-closed APIs, rate limits, and safe degradation.

Next.js 16
React 19
Supabase
PostgreSQL
Gemini
Vitest

Recognition

Proof that holds up outside the repo.

A company, a scholarship, a paper, a copyright, and national programs, each one independently verifiable.

Co-founder · 2025–present

Plus The Site

Co-founded plus., a digital AI agency where one team handles AI chatbots, branding, apps, and CRM instead of five separate vendors. I lead across design, development, and marketing, and the agency was selected for the SIAP 2025 acceleration program by Kementerian Ekonomi Kreatif.

Visit plusthe.site

Build smarter brands. Faster. AI + human creativity · EN / ID

Scholarship · 2024–2026

Bank Indonesia Scholarship

Full academic scholarship for excellence and leadership potential, covering my final two years at Telkom University.

Research · IEEE Xplore

First-author publication

A study on synthetic IoT data generation, indexed internationally with me as first author.

IP rights · Kemenkumham RI

Registered software copyright

“Tel-U Great Planner Academic”, Reg. No. 000925616, registered 2025.

Government program · 2025

SIAP acceleration participant

Selected for Kemenekraf’s national acceleration program for creative-economy ventures.

National finalist · Telkom Indonesia

Innovillage Top 163

An IoT flood early-warning system that made the national finals two years running, in 2023 and 2024.

Security research · HackerOne

TikTok vulnerability report

Found a client-side audio-restoration flaw independently and disclosed it responsibly.

Olympiad · Astronomy

OSN provincial finalist

National Science Olympiad finalist at provincial level.

Professional certification

BNSP digital marketing

Nationally certified competency in digital marketing practice, 2024.

Facilitator · Dicoding Indonesia

Google Skills Arcade 2026

Guiding a cohort through Google Cloud learning paths (Jul–Sep 2026), backed by my own Gen AI, LLM, and Responsible AI credentials on Google Skills.

Process

How a change earns the word “done.”

The same five steps sit behind every test count on this page, from the flagship builds to the smaller systems further down.

I use Claude Code every day. But traces, diffs, tests, builds, security checks, and human review decide when work ships, not the AI.

01
Observe
Reproduce the failure and capture the smallest trace, log, or test that shows it.
02
Decompose
Split analysis, frontend, backend, data, and QA work wherever running them in parallel actually helps.
03
Implement
Keep ownership boundaries explicit and change only what the evidence supports.
04
Challenge
Attack permissions, error paths, security assumptions, and regression risk before anyone else has to.
05
Verify
Run the tests and the build, walk the real user path, and write down what’s still uncertain.

Human integration gate

One accountable owner.

Specialist agents can chase separate questions in parallel. I reconcile what they find against the repository and own the final call. When they disagree, I go get more evidence.

01Scoped roles
02Clear acceptance criteria
03Dependency-aware sequencing
04QA before “done”

Claude Code
Next.js
Supabase
Python
Flutter
GitHub Actions

More builds

Six more systems. Same standard.

Smaller in scope than the flagships, built with the same habit: a peer-reviewed paper, public repositories, and running apps.

Where a project has a limit, like a retired deployment or credit shared with a team, the card says so up front.

First-author research · IEEE Xplore

Synthetic IoT data for smart agriculture

A comparison of Gaussian mixture models and Monte Carlo simulation for generating synthetic sensor data in durian-cultivation smart-farming IoT systems.

Monte Carlo
GMM
Smart farming
IoT data

Read on IEEE Xplore

Public repository · 24 tests

Shariah Trading Assistant

A multi-agent research and analysis system with scoped roles, careful evidence handling, and explicit QA gates.

Multi-agent
Orchestration
QA gates

Read the source

Team build · project lead

Protein RADAR

A Next.js and PostgreSQL dashboard pulling in more than twenty public data sources, with authentication, RBAC, and audit logging. I led the project within a documented team.

Next.js
Prisma
PostgreSQL
RBAC

Open the live app

Production IoT · Desa Mukai Tengah, Kerinci

Flood early-warning system

A solar-powered IoT platform from Tim Rewana that monitors water clarity and flood risk for roughly 300 households, with sensor integration and a real-time web dashboard.

C++
IoT
Embedded
PHP

Backend engineering · computer vision

AI surveillance dashboard

Backend services, data pipelines, and REST APIs for real-time video analytics, alert management, and uptime monitoring, integrating 10+ computer-vision models over WebSocket.

Python
REST API
WebSocket
Data pipelines

Computer vision · archived

Qoffea

A coffee-bean grading workflow with Flask, YOLO inference, confidence filtering, non-maximum suppression, API tests, and container setup.

The deployment is retired, so this links to the source only.

Python
Flask
YOLO
Docker