AI & Software

I build AI systems that ship to production.

I'm focused on artificial intelligence and machine learning, from model pipelines and computer vision to the backend systems that put them in front of real users. I like owning a problem end to end, from the data and the model to the API that finally serves it.


6

Projects shipped across ML, systems and web

2

Engineering roles, in fullstack and robotics

3

National & international competition finals


About

From raw data to shipped systems

I approach every problem with a product mindset. A model is not done when it's accurate, it's done when it's deployed, monitored, and used. That mindset carries across my work in machine learning, backend engineering, and the embedded systems underneath both.

Portrait of Ariel Cornelius Sitorus

Expertise

AI & Machine Learning

Pipelines from raw features to a model in production: ensemble classifiers, demand forecasting, and real-time computer vision detection. Comfortable in both the notebook and the API that serves the result.

PyTorchscikit-learnXGBoostOpenCVYOLO

Software & Data Systems

Full-stack products and the data layer underneath them: APIs, relational databases, and containerized deployments, plus the pipelines that turn raw data into something a model or a dashboard can use.

PythonGoFastAPIPostgreSQLDocker

Systems & Robotics

Kernels, drivers and flight systems, comfortable a level below the abstraction most engineers stop at, from interrupt handling to autopilot integration on a live airframe.

CAssemblyMAVLinkRaspberry Pi

Projects

Proactive Phishing URL Detector

Proactive Phishing URL Detector

ML pipeline for classifying phishing URLs purely from raw URL structure. Uses 70+ hand-crafted lexical features combined with TF-IDF character n-gram SVD embeddings fed into a weighted ensemble.

PythonXGBoostLightGBMCatBoost
FMCG Market Demand Forecasting

FMCG Market Demand Forecasting

Forecasting project on 1M+ FMCG transactions across 15 SKUs. Built 30+ features and benchmarked models for pricing strategy, inventory planning, and product cannibalization analysis.

Pythonscikit-learnProphetstatsmodels
Online Gambling Detector

Online Gambling Detector

Chromium extension that scans webpages for gambling content using Aho-Corasick, KMP, Boyer-Moore, Rabin-Karp, and fuzzy matching. Uses OCR for embedded graphics and mutation observers for dynamic page updates.

TypeScriptViteChart.jsTesseract.js
E-Library Search Engine

E-Library Search Engine

Digital library web app with text, document upload, and book cover image search. Text search runs Latent Semantic Analysis and TF-IDF, while cover search uses PCA-based image similarity.

GoTailwind CSS
DOM Tree Traversal & Visualization Platform

DOM Tree Traversal & Visualization Platform

Web platform for querying and visualizing the DOM tree of any URL. Parses raw HTML with a custom tokenizer and traverses it with BFS or DFS through a custom CSS selector engine.

TypeScriptExpressDockerAzure
32-bit x86 Kernel Implementation

32-bit x86 Kernel Implementation

32-bit OS kernel built from scratch in C and x86 Assembly. Covers GDT/IDT setup, PIC-managed interrupt handling, paging-based virtual memory, round-robin scheduler, and FAT32 driver.

CAssemblyx86QEMU

Papers


“Build things that matter.”
Ariel Cornelius Sitorus

Experience

ML and Robotic Software Developer

Jan 2025 to Present

Akasadipha

Akasadipha is part of Aksantara ITB, focused on the development of UAV systems for long endurance and low altitude missions.

  • Developed a real-time hotspot detection system using OpenCV and YOLO-based object detection.
  • Implemented communication between onboard systems and the flight controller using MAVLink.
  • Designed and integrated a payload drop mechanism connected to the autopilot (Matek) and controlled via a Raspberry Pi.
  • Utilized Mission Planner for telemetry monitoring and mission data integration.
OpenCVYOLOMAVLinkRaspberry Pi

Software Engineer Associate

Feb 2025 to Present

Fourthirty RC

Fourthirty RC is a community-based sports initiative focused on running, combining regular athletic activities with the development of sportswear products such as apparel and hydration gear.

  • Built a full e-commerce platform with product catalog, variant system, session-based cart, and real-time JNE shipping cost calculation via RajaOngkir API.
  • Built and refined key interface elements such as navigation bars, dashboard views, and reusable UI components.
  • Owned the full design-to-production pipeline by designing in Figma and building responsive Blade templates.
  • Contributed to the design of sports apparel, including running shirts and activewear.
Laravel 9MySQLPHP 8FigmaBlade

Achievements

The Great Data Heist

National Finalist

2025

ICPC Asia Jakarta

International Contestant

2024

Kompetisi Robot Terbang Indonesia (KRTI)

National Finalist & Best Spirit Award

2024

Let's talk AI, systems, or your next hard problem.

Open to internships, collaborations, and technical conversations. Reach me directly at arielsitorus504@gmail.com, or find me here.