Irfan Fadhullah

Muhamad Irfan Fadhullah

AI Engineer & Researcher

Building intelligent systems with Computer Vision, LLMs, and 3D Deep Learning. Currently at Realteeth, Seoul.

LinkedIn GitHub Medium

About Me

AI Engineer & Researcher with expertise in Computer Vision, Large Language Models, and Deep Learning. Currently building AI-powered dental analysis systems at Dentalverse Inc. in Seoul, South Korea.

My work spans from developing vision-language models for medical imaging to creating autonomous coding agents. I hold a Master's degree from Sejong University (GPA: 4.00/4.50) and a Bachelor's from Telkom University (GPA: 3.79/4.00, Cum Laude).

4+
Years in AI/ML
10+
Research Projects
13
Published Articles
144x
Faster OCR Pipeline

Computer Vision

Object detection, tracking, segmentation, and video captioning using YOLO, ViT, and custom architectures.

Large Language Models

Building AI agents with LLMs, RAG, and PEFT. Fine-tuning with LoRA and quantization techniques.

Medical AI & 3D Dental

Dental X-ray analysis and disease detection at Dentalverse. 3D tooth segmentation, completion, crown generation, and margin line generation at Realteeth.

Experience

Realteeth

Dec 2025 — Present
AI Engineer
Seoul, South Korea
  • 3D tooth segmentation on point cloud data for precise dental analysis.
  • 3D tooth completion — reconstructing missing tooth geometry from partial scans.
  • 3D single crown generation for dental prosthetics design.
  • Margin line generation on 3D point cloud data for crown fitting.

Dentalverse Inc.

Aug 2024 — Dec 2025
AI Researcher
Seoul, South Korea
  • Develop deep learning models for FDI tooth numbering, disease detection (calculus, periodontitis, abscess, cyst), and anatomical segmentation.
  • Build CV systems for multi-angle FDI numbering, caries detection, prosthesis identification, and arch classification.
  • Design agentic models integrating VLMs with specialized expert models for complex dental case management.
  • Develop generative models (diffusion, GAN) for visualizing post-treatment changes.

Sejong University

Mar 2022 — Jul 2024
AI Researcher & Teaching Assistant
Seoul, South Korea
  • Collaborated on 4 key areas of CV and ML on the ICT 4.0 Project by the Korean Government.
  • Developed vision-based wireless communication algorithms using YOLO and ViT. Outperformed prior work on 20/21 scenarios by 3%+.
  • Built a language-driven computer task agent with LLM + OCR + RAG, achieving 22%+ improvement using 8-bit quantization.
  • Reproduced notable papers on object detection, video captioning, and vision-based communication.

PT. Simas Jiwa Insurance

Nov 2020 — Feb 2022
Data Scientist
Jakarta, Indonesia
  • Built personalized recommendation systems, AI search engine, and document categorization tools.
  • Developed LSTM-based customer service chatbot with faster response than human agents.
  • Implemented OCR and KYC approval model, reducing processing from 1 day to 10 minutes (144x faster).

IYKRA Data Fellowship

Apr 2020 — Aug 2020
Data Scientist Awardee
Jakarta, Indonesia
  • Trained in Data Visualization, Statistics, ML, Big Data Tools, and more.
  • Ranked Top 3 Best Participant and Best Project Solver with TEDx Jakarta.
  • Conducted impact measurement using Social Network Analysis.

Education

Sejong University

M.S. Intelligent Mechatronics & Convergence Drone
Seoul, South Korea · Graduated Aug 2024
GPA: 4.00 / 4.50
  • Lab: Vision Language Intelligence
  • Thesis: A Language-Driven Agent for Computer Tasks
  • Built agents automating computer tasks using LLM/VLM + OCR + RAG

Telkom University

B.Eng. Telecommunication Engineering
West Java, Indonesia · Graduated Mar 2020
GPA: 3.79 / 4.00 — Cum Laude
  • Lab: Mobile Communication (Cellular Network & IoT)
  • OPES Scholarship recipient
  • Published research on GSM/LoRa signal interference

Research & Projects

Selected work in AI, computer vision, and language models

UNet Zoo — PyTorch Library

A unified PyTorch library for UNet architecture variants. Provides standardized implementations for semantic segmentation tasks.

PyTorchSegmentationOpen Source
GitHub Article

Coding Computer Agent

An agent using dual LLMs (Phi 3 Mini + DeepSeek Coder-2) for coding tasks. Fine-tuned with LoRA on a custom dataset.

LLMLoRAAgent
GitHub

Language-Driven Computer Agent

Autonomous agent using Llama 3 + OCR + RAG for computer tasks. 22%+ success rate improvement over prior work with 8-bit quantization.

Llama 3RAGOCRPEFT
GitHub

Vision-Based Blockage Detection

Deep learning model for blockage detection in wireless communication using Faster-RCNN + ViT + LSTM. Improved 20/21 scenarios by 3%+.

Faster-RCNNViTLSTM
GitHub

Video Captioning (Seq2Seq)

Reproduced seq-to-seq video captioning with LSTM + CNN. Proposed model achieved 20% BLEU score improvement over baseline.

LSTMCNNNLP
GitHub

Vehicle Tracking in Adverse Weather

Extended CenterTrack for multi-weather tracking with adaptive classification. 2%+ MOTA improvement.

CenterTrackTrackingWeather
GitHub

Continual Learning for Dynamic Inputs

Elastic Weight Consolidation + ResNet for knowledge retention. 30% accuracy improvement (50% → 80%+) vs baseline.

EWCResNetContinual Learning
GitHub

Awesome Indonesian LLM Dataset

Curated collection of datasets for Indonesian language LLMs — a community resource for Indonesian AI development.

LLMDatasetIndonesian NLP
GitHub

Hybrid Beamforming with Vision

Vision-based hybrid beamforming in vehicular networks using YOLO + ViT. 50.17% improvement in achievable rate.

YOLOViTWireless
GitHub

Blog & Articles

Writing about AI, machine learning, and data science on Medium

Jul 2025

UNet Zoo: A PyTorch Library for UNet Variants

A unified library for UNet variants with standardized implementations for various segmentation tasks.

Read on Medium
Jul 2025

🇮🇩 Awesome Indonesian LLM Dataset

A curated collection of datasets for Indonesian LLMs — a game-changer for Indonesian AI development.

Read on Medium
Mar 2025

Survey on Text-to-Speech Works

Comparative analysis of modern TTS models and their architectures.

Read on Medium
Mar 2025

Paper Review: Contextual Convolutional Networks

In-depth review of the ICLR 2023 paper on contextual convolutional networks.

Read on Medium
Mar 2025

The ABCs of CV Segmentation

Object Detection vs. Instance vs. Semantic vs. Panoptic segmentation explained.

Read on Medium
Oct 2020

Machine Learning 101

A beginner-friendly explanation of machine learning basics and common algorithms.

Read on Medium
View All Articles on Medium

Skills & Expertise

Languages & Tools

PythonJavaC++ PyTorchTensorFlowLangChain SQLSparkBigQuery

AI & Research Fields

Deep LearningComputer Vision Object Detection & TrackingImage & Video Captioning Large Language ModelsRAG Semantic SegmentationGenerative Models 3D Generation & Completion3D Point Cloud Segmentation GANDiffusion Models

Cloud & Infrastructure

Google Cloud PlatformAzure DockerLinux

Key Certifications

TensorFlow: Advanced TechniquesNLP Specialization Deep Learning SpecializationGANs Specialization ML with TensorFlow on GCPAdvanced ML on GCP

Get in Touch

I'm always open to discussing new opportunities, research collaborations, or interesting AI projects.