Sep 2026 – Present
Solution Architect
Fujitsu
Jakarta
Full-time
Hi, I'm Fadhel Haidar
AI & Machine Learning Engineer
Currently:
Solution Architect @ Fujitsu | AI/ML Mentor @ PT Dibimbing Digital Indonesia
My work spans traditional machine learning, computer vision, NLP, and agentic AI systems. I take AI projects from the first experiment to the infrastructure they run on: training and evaluation, deployment, and inference optimization for servers and edge devices.
3+
Years experience
6
Companies
Experience
6 records. What shipped while I was there.
Jul 2024 – Aug 2026
Data Scientist
PT United Tractors Tbk
Jakarta
Full-time
- Built an agentic RAG system over vector and SQL databases: sales data preparation dropped from about 5 days to 15 minutes.
- Shipped a multimodal LLM PDF extraction system for 30-page monitoring documents, cutting extraction from 30 minutes to 5.
- Developed an MLOps system with MLflow to automate batch inference workflows.
Dec 2024 – Present
AI/ML Mentor
PT Dibimbing Digital Indonesia
Jakarta
Freelance
- Developed teaching materials aligned with the bootcamp's standard curriculum.
- Taught classes of up to 30 students through interactive lessons and hands-on exercises.
- Assessed student assignments and gave feedback on their work.
Mar 2024 – Jul 2024
Artificial Intelligence Engineer
PT Asa Ren Global Nusantara
Jakarta
Full-time
- Built a patient analysis system on an agentic RAG framework retrieving from multiple APIs, helping patients understand their conditions.
- Created an OCR system using multimodal LLMs to extract varied electronic health record documents.
- Mapped biomedical terms to standardized formats with a semantic search system.
Aug 2023 – Mar 2024
Artificial Intelligence Engineer
PT Wide Technologies Indonesia
Jakarta
Full-time
- Built a face analysis model at 85% face-shape classification accuracy (EfficientNet) for a glasses recommendation system.
- Built a skin analysis service for a telemedicine app: InceptionV3 for wrinkles and dark circles, DETR for acne detection, VGG16 for skin type.
- Shipped an eKYC API (MTCNN, ArcFace, multi-similarity verification) that improved onboarding speed by 50% over the previous liveness system.
May 2023 – Sep 2023
Artificial Intelligence Engineer
GAIA Dental Studio
Surabaya
Part-time
- Built a dental disease detection API (YOLOv8, VGG16) at 70% mAP and 89% precision for early screening.
- Built an ocular disease classification API with VGG16 at 89% precision for preliminary diagnosis.
- Automated invoice extraction with Google Vision OCR and GPT-3.5, cutting manual data entry and validation time.
Education
3.71
GPA / 4.00
Brawijaya University
Bachelor of Computer Engineering
Aug 2019 – Jan 2023
Malang, Indonesia
Skills
Programming & Frameworks
Python
TensorFlow
PyTorch
FastAPI
LangChain
LlamaIndex
Transformers
Unsloth
Databases
PostgreSQL
MongoDB
BigQuery
Deployment & MLOps
Docker
TensorRT
OpenVINO
MLflow
vLLM
Projects
Built end to end. Where the code can't be shared, a video demo stands in.
Library
recon-graphrag
$ pip install
Open Source
NLP
Featured
Recon-GraphRAG
A domain-agnostic GraphRAG SDK built on Neo4j and neo4j-graphrag. It follows the Microsoft GraphRAG approach with hierarchical community detection and supports Local, Global, and DRIFT search for question answering across knowledge graphs.
GraphRAG SDK · Community Detection · Hybrid Search
Neo4j
Graph Data Science
Python
$ pip install git+https://github.com/FadhelHaidar/recon-graphrag.git
NLP
Featured
Agentic Chatbot System - Salesman Knowledge Expert
This is a demo for Agentic Chatbot System using Llamaindex and Vector DB. The system can perform sales knowledge analysis and generate a report for the user through Text2SQL. Also the chatbot is equipped with a vector DB to store sales knowledge and storing SQL related information to help performing Text2SQL.
Chatbot · LLM
Llamaindex
Vector DB
Transformers
Computer Vision
Robotics
Featured
Human Tracking System for Smart Wheelchair
The system Leveraging YOLOv5 and byte track to create auto navigation by following human in front of the smart wheelchair captured by web camera. The system is embedded to Nvidia Jetson TX2.
Object Detection · Object Tracking
Nvidia Jetson
Arduino
Pytorch
NLP
Agentic Chatbot System - Fraud Analysis
This is a demo for Agentic Chatbot System using Langgraph and Vector DB. The system can perform analysis on historical fraud data and generate a report for the user through Text2SQL. Also the chatbot is equipped with a vector DB to store fraud theory knowledge and storing SQL related information to help performing Text2SQL.
Chatbot · LLM
Langgraph
Vector DB
Transformers
Computer Vision
Skin Analysis AI
A feature that uses DETR object detection model with a custom dataset to identify acne problems, InceptionV3 for wrinkle classification, mediapipe for facial landmark detection, and VGG16 for skin type classification. The same method for wrinkle detection is applied to detect dark circles under the eyes.
Object Detection · Landmark Detection · Image Classification
Flask API
Tensorflow
Computer Vision
Facial Ratio Measurement AI
Using EfficientNetV2, the system can classify the shape of the face. It also calculates the width-to-length ratio of the face using Euclidean Distance to measure the distance between facial landmarks.
Object Detection · Landmark Detection · Image Classification
Flask API
Tensorflow
Computer Vision
E-KYC Face Recognition System and OCR ID Card Reader
The system can perform face recognition using FaceNet and read ID card information using Transformer OCR. The system is designed for electronic Know Your Customer (e-KYC) processes, allowing for efficient and accurate identity verification.
Object Detection · Optical Character Recognition · Face Recognition
Flask API
Tensorflow
Transformers
OpenVINO
Computer Vision
Robotics
Room Name Recognition System
The system contained object detection to detect the plate using YOLOv5 and read/recognize the plate character using EasyOCR, the object captured using web camera. The system is embedded to Nvidia Jetson TX2.
Object Detection · OCR
Nvidia Jetson
Arduino
Pytorch
Research
Published work in AI and machine learning.
Computer Vision
Object Detection
OCR
2023 · Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer (J-PTIIK)
Deteksi Plat Nama Ruangan untuk Kendali Kursi Roda Pintar menggunakan YOLOv5 dan EasyOCR berbasis TX2
Muhammad Fadhel Haidar1, Fitri Utaminingrum2
Penyandang disabilitas di Indonesia mencapai 5% dari jumlah populasi penduduk Indonesia termasuk tuna daksa dan tuna netra. Berdasarkan masalah dialami oleh penyandang disabilitas atas keterbatasan ruang gerak, maka diperlukan pengembangan alat bantu yang dapat memberikan manfaat bagi penyandang disabilitas. Kursi roda merupakan alat bantu yang dapat membantu penyandang disabilitas dalam beraktivitas. Kursi roda elektrik memiliki keunggulan berupa kemudahannya dalam dikendalikan dan bermanuver, kemampuannya dalam melewati bidang miring, serta lebih mudah untuk dibawa ke tempat yang relatif jauh. Pada penelitian ini bertujuan untuk mengembangkan sistem yang dapat digunakan untuk mendeteksi plat nama ruangan secara otomatis pada kursi roda pintar. Pada sistem ini objek papan nama ruangan yang ditangkap oleh kamera akan dijadikan sebagai tujuan dari pergerakan kursi roda secara otomatis. Objek tangkapan kamera akan diolah menggunakan metode YOLOv5 dan EasyOCR. Penelitian ini dapat membuktikan bahwa nilai akurasi terhadap plat nama ruangan dengan menggunakan metode YOLOv5 mendapatkan nilai akurasi sebesar 60% sedangkan 100% untuk metode EasyOCR. Hal ini memungkinkan kemampuan integrasi sistem deteksi objek dan pengenalan karakter ruangan pada sistem autonomous terhadap peningkatan nilai PWM (pulse width modulation) yang dapat diartikan jika sistem tidak dapat mengenali karakter plat nama ruangan yang mengakibatkan kursi roda terus berjalan hingga menemukan objek yang dicari.
Methods
YOLOv5
EasyOCR
Computer Vision
Deep Learning
2024 · Journal of Materials Research and Technology
A convolutional neural network-VGG16 method for corrosion inhibition of 304SS in sulfuric acid solution by timoho leaf extract
Femiana Gapsari, Fitri Utaminingrum, Chin Wei Lai, Khairul Anam, Abdul M. Sulaiman, Muhamad F. Haidar, Tobias S. Julian, Eno E. Ebenso
A corrosion inhibition test, coupled with a quantification of in-situ H2 evolution, can be used to evaluate an organic inhibitor such as Timoho leaf extract (TLE). TLE is a biodegradable and effective corrosion inhibitor because of its potential to protect 304SS against sulfuric acid. TLE corrosion inhibitor was studied through systematic electrochemical experiments and morphological characterization, with a concentration range of 0–6g L−1. Convolutional Neural Network (CNN)-VGG16 was one of the machine learning approaches used to classify and predict physical changes in hydrogen gas bubbles. Constituents of the TLE and 304SS surfaces were analyzed by FT-IR and UV–Vis tests. The results suggested that 3 g L−1 TLE inhibitor was able to reduce the corrosion rate by 99.37 %. The TLE's inhibition mechanism on 304SS was mixed adsorption and mixed type inhibitor that followed the Isothermal Freundlich Equation. The prediction model by CNN-VGG16 for corrosion tests at varied inhibitor doses was 96% accurate. SEM tests revealed that TLE constituent adsorption on the 304SS surface had a smooth surface morphology with few degraded spots.
Methods
VGG16
CNN
Get in touch
Open to freelance AI projects, part-time and contract work, research collaborations, and full-time roles.
Open to opportunities
Project Development
Machine learning, computer vision, and NLP systems end to end: model training and optimization, deployment to embedded devices or production servers, and the REST API, database, and authentication layer around them.
Technical Advisory
System architectures for ML products, code reviews, and guidance for teams taking AI/ML from prototype to production.
Ways to work together
Freelance projects
Short to medium-term projects with clearly defined scope and deliverables.
Collaborations
Co-development on AI and ML projects, research partnerships, or open-source contributions.
Part-time and contract
Part-time roles or contract work for companies seeking AI expertise.
Full-time roles
Full-time positions at companies working on AI and ML. Open to remote, hybrid, or on-site roles.
Send a short note about your project or opportunity, and I'll take it from there.
What to include
- Brief description of your project or opportunity
- Timeline and expected deliverables
- Budget range (if applicable)
- Any specific requirements or constraints
Response time
I typically respond within 1-2 business days and can usually schedule a call the same week.
Have a project in mind?
Open to freelance, contract, and full-time roles in AI and machine learning.
Get in touch