Physical AI Engineer

Muhammad Miftah Faridh

Building intelligent systems at the intersection of AI, robotics, and edge computing. From LLM fine-tuning and smart braille printers to autonomous marine vehicles.

M.S. Electronics Engineering, Kookmin University — Currently at orinu.ai, Seoul

Get in Touch Google Scholar
7+
Publications
5
Int'l Awards
8
Gov. Projects
20+
Open Source
Muhammad Miftah Faridh
About

Engineer, Researcher
& Problem Solver

AI Developer and Robotics Engineer with hands-on experience in LLM fine-tuning (T5, Gemma, LLaMA with LoRA, QLoRA, and RAG) and real-time robotic systems using ROS 2 and NVIDIA Jetson platforms. Proven track record in applied research, publishing papers in IEEE Access and MDPI Sensors, and building intelligent systems across industries.

From autonomous marine vehicles at University of Indonesia's Robotics Team to 5G-connected drones at XL Axiata, from Optical Camera Communication research at Kookmin University to Smart Braille Printers at orinu.ai — I bridge AI and robotics to solve real-world problems.

🎓

M.S. EE, Kookmin University

Seoul, South Korea (2022 – 2024)

🎓

B.S. Computer Engineering

University of Indonesia (2018 – 2022)

⚙️

Physical AI Engineer

orinu.ai — Hwaseong, KR

📍

Based in Seoul

Open to global opportunities

Expertise

Core Competencies

Bridging AI research and engineering across robotics, edge computing, communications, and intelligent systems.

01

LLM & Generative AI

Fine-tuning large language models (T5, Gemma, LLaMA) with LoRA, QLoRA, and RAG pipelines. Building translation engines and conversational AI for production deployment.

02

Robotics & ROS 2

Real-time robotic systems on Jetson platforms. Robotic arms, autonomous marine vehicles (USV/AUV), and drone systems with integrated computer vision pipelines.

03

Edge AI & Deployment

8+ production Edge AI systems on NVIDIA Jetson. TensorRT INT8/FP16 optimization achieving 67% model compression. Full-stack: FastAPI + React + WebSocket streaming.

04

Deep Learning

CNN, LSTM, GCN, and hybrid architectures for signal processing, image recognition, and time-series forecasting. Published research on solar energy prediction and OCC systems.

05

Optical Camera Comm.

2D MIMO OCC systems, CSMA/CA MAC protocol design for optical communication, and ultra-high-rate OCC with precision localization. Published in IEEE Access.

06

IoT & Energy Systems

Intelligent IoE platforms, sensor integration with LoRa mesh networks, solar energy forecasting with GCN-BiLSTM models, VPP ESS platforms, and V2X communication.

Publications

Research Papers

Peer-reviewed publications in international journals, conferences, and technical competitions.

2025
IEEE Access

Implementation of CSMA/CA-Based MAC Protocol for Optical Camera Communication

M. M. Faridh et al.

→ 2023
Sensors (MDPI)

Intelligent IoT Platform for Multiple PV Plant Monitoring

M. M. Faridh et al.

→ 2023
Sensors (MDPI)

Design and Implementation of a 2D MIMO OCC System Based on Deep Learning

M. M. Faridh et al.

→ 2023
ICUFN

Intelligent IoT Platform for PV Power Plant Monitoring and Control

M. M. Faridh et al.

→ 2023
ICAIIC

Forecasting Solar Energy Production using a Hybrid GCN-BiLSTM Model

M. M. Faridh et al.

→ 2021
RoboNation

MAKARA MH4 XX — Autonomous Surface Vehicle

AMV UI Team

→ 2019
RoboNation

Makara 09 Mark II — Autonomous Surface Vehicle

AMV UI Team

→
Portfolio

Projects & Open Source

AI/ML research, Edge AI deployments on Jetson, IoT systems, and full-stack applications.

AI & Deep Learning Research 2 projects
Deep Learning / Energy Forecasting / Published

PV Solar Power Forecasting API

Next-day photovoltaic solar power prediction across 10 South Korean sites. Flask REST API with 6 deep learning architectures trained on Korea Meteorological Administration weather data.

Models: BiLSTM, Stacked BiLSTM, LSTM, ConvLSTM, RNN — trained on 12 weather features (soil temp, solar radiation, cloud cover, etc.) with 24-hour sliding window producing 24-hour forecasts. MinMax normalization with tanh activation.

Result: BiLSTM variants achieved lowest prediction error. Automated daily predictions at 10:00 and 16:00 KST. Published in Sensors 2023 (MDPI) and ICUFN 2023.
TensorFlow Flask BiLSTM ConvLSTM MariaDB
10Solar sites
6DL models
12Features
View on GitHub
Deep Learning / Time-Series

Smart Home Energy Forecasting

Comparative study: ARIMA/SARIMA/SARIMAX vs LSTM/BiLSTM/GRU/RNN on 503,910 smart meter records with weather features.

Key finding: LSTM Multivariate achieves MSE=0.022, MAE=0.110 — significantly outperforming classical methods. Extended to 17 South Korean regions with BiLSTM achieving R²=0.9549 on Busan data.
TensorFlow BiLSTM GRU ARIMA Jupyter
View on GitHub
LLM / Agents

LangGraph Chatbot Playground

Experimental conversational AI with LangGraph — stateful multi-turn dialogue, agent workflows, and LLM orchestration using OpenAI/Anthropic/OpenRouter APIs.

Python LangGraph LangChain LLM
View on GitHub
Edge AI on NVIDIA Jetson 8 projects
Edge AI / Deployment Hub

Edge AI Deployment Platform

Central deployment hub for all 8 edge AI projects. TensorRT INT8/FP16 model compression achieving 67% total size reduction (482 MB to 152 MB). Includes all compressed models, inference scripts, SCP/USB transfer utilities, and shared Python library with TensorRT inference engine, async video capture, WebSocket streaming, and Docker base images.

TensorRT Docker Jetson Nano FastAPI React WebSocket CUDA
67%Compression
152MBTotal size
8AI projects
View on GitHub
Computer Vision / 80 Classes

Object Detection

Real-time 80-class COCO detection. Full stack: FastAPI backend, React/TypeScript frontend, WebSocket streaming. TensorRT FP16 targeting 20-25 FPS at 416x416 on Jetson Nano.

YOLOv8TensorRTFastAPIReact
View on GitHub
Biometrics / Anti-Spoofing

Face Recognition Attendance

InsightFace buffalo_s with RetinaFace + ArcFace recognition + MiniFASNet anti-spoofing. Multi-angle enrollment (5 angles). Full stack with PostgreSQL. INT8 compressed: 158 MB to 40 MB. 15-20 FPS.

InsightFaceArcFacePostgreSQLFastAPI
View on GitHub
Safety / Zone Compliance

PPE Detection

Multi-PPE detection (Hardhat, Vest, Glasses, Gloves, Mask) with zone-based compliance for Construction/Manufacturing/Chemical zones. YOLOv8n with TensorRT INT8 (3.4 MB model).

YOLOv8TensorRT INT8Zone Detection
View on GitHub
OCR / Vehicle Tracking

License Plate Recognition

YOLOv8 plate detection + EasyOCR/PaddleOCR. SORT vehicle tracking. Indonesian plate format validation. Watchlist monitoring with alerts. Full stack with PostgreSQL + React.

YOLOv8EasyOCRSORTReact
View on GitHub
Surveillance / Anomaly

Security Analytics

Motion detection (MOG2/KNN), polygon zone-based intrusion detection, and loitering detection with time tracking. YOLOv8n person detection + DeepSort tracking. INT8/FP16 compressed.

YOLOv8DeepSortMOG2/KNN
View on GitHub
People Counting / Traffic

People Analytics

Line crossing counter, zone occupancy monitoring, traffic flow analysis. HOG fallback when no ML model available. JSON statistics output. 25-30 FPS with TensorRT on Jetson.

YOLOv8HOGLine Crossing
View on GitHub
Gesture / HRI

Hand Gesture Recognition

MediaPipe 21-landmark hand tracking with 9 gesture classes (FIST, OPEN_PALM, THUMBS_UP, etc.). Custom ML classifier training pipeline. Action mapping with cooldown. 15-20 FPS on Jetson.

MediaPipeCustom MLOpenCV
View on GitHub
Pose Estimation / Fitness

Exercise Tracker

Push-up, squat, plank tracking using MediaPipe/YOLOv8-pose/MoveNet. Rep counting via joint angle calculation. Real-time form analysis with corrective feedback.

MediaPipeYOLOv8-poseMoveNet
View on GitHub
IoT & Security Systems 2 projects
IoT / LoRa Mesh / Monitoring

Server Room Monitoring

Multi-sensor IoT system: Arduino nodes with DHT22, dust (PM2.5/PM10), vibration, and sound sensors. LoRa wireless mesh to central receiver. Python backend logging to MariaDB with Telegram bot alerts on anomalies.

ArduinoLoRaMariaDBTelegram
View on GitHub
Computer Vision / Alerts

Intruder Detection

Real-time face detection using OpenCV DNN (Caffe SSD model) with automatic Telegram photo alerts. Configurable confidence threshold and frame skip rate. Non-blocking async notification delivery.

OpenCV DNNCaffe SSDTelegram Botasyncio
View on GitHub
Web Applications 4 projects
Speech-to-Text / Google Cloud

VTT — Voice-to-Text

Indonesian speech-to-text using Google Cloud Speech-to-Text API. FLAC audio processing via FFmpeg. Async long-running recognition optimized for Bahasa Indonesia (id-ID). Google Cloud Storage for audio files.

Laravel 5.6Google CloudVue.jsFFmpeg
View on GitHub
NLP / Social Platform

TweetBullyingDetector

Cyberbullying detection and reporting platform for Twitter/X. Keyword-based tweet search via Twitter API v1.1, community reporting system, and educational materials on cyberbullying awareness.

Laravel 7Twitter APIPHPMySQL
View on GitHub
Management / Livewire

VeTracker — Vehicle Manager

Vehicle inventory management with CRUD operations, image upload, and multi-user team support with Jetstream 2FA authentication. Reactive UI with Livewire components.

Laravel 8LivewireTailwind CSS
View on GitHub
Tournament / Events

Brocket — Bracket Manager

Tournament management with auto-generated randomized brackets, podium/medal tracking, public bracket sharing, and email notifications. Jetstream auth with 2FA. Team project (7 members).

Laravel 8LivewireTailwind CSS
View on GitHub
Honors

Awards & Competitions

Recognized in international and national robotics competitions representing Indonesia.

🥇
1st Place

ASEAN MATE ROV — Explorer Category

International underwater robotics competition

🥉
3rd Place

12th AUVSI International Roboboat

Miami, FL, USA · 2019

🥉
3rd Place — Sensor Optimization

14th AUVSI International Roboboat

2021

🥉
3rd Place

KKCTBN 2019

Kontes Kapal Cepat Tak Berawak Nasional, Indonesia

🏆
Best Poster

KKCTBN 2019

Kontes Kapal Cepat Tak Berawak Nasional, Indonesia

Research

Government-Funded Projects

Contributed to major national research projects at Kookmin University, funded by Korean government ministries.

MOTIE

Renewable Energy Grid Stabilization

2022 – 2024

MSIT

5G Multi-Band Industrial Small Cell Development

2022 – 2024

MSIT

Intelligent IoE Data Research Platform

2022 – 2024

MSIT

V2X Quantum AI Triple Sensor OCC System

2022 – 2024

MSS

VPP ESS Service Platform Development

2022 – 2024

MSIT

Ultra High Rate OCC & Precision Localization

2022 – 2024

Career

Professional Journey

Oct 2025 — Present

Physical AI Engineer

orinu.ai · Hwaseong, South Korea

  • LLM fine-tuning for translation & Retrieval-Augmented Generation (RAG) projects
  • Smart Braille Printer development using robotic systems
  • Robotic arm development with ROS 2 on Jetson platforms
Sep 2022 — Aug 2024

AI & IoT Researcher

Kookmin University · Prof. Jang Yeong Min's Lab, Seoul

  • Engineered LED detection and tracking systems for adverse weather conditions
  • Built time-series forecasting models to predict PV power generation using weather data
  • Published 2 international journals and 2 conference papers within 2 years
  • Contributed to 8 government-funded research projects (MOTIE, MSIT, MSS)
Feb 2022 — Jun 2022

Internet of Things Intern

PT. XL Axiata Tbk · Jakarta, Indonesia

  • Drone project development with 5G connectivity
  • Implemented computer vision on drones with 100% success rate
Sep 2021 — Feb 2022

IT Business Analyst

PT. Infomedia Nusantara · Jakarta, Indonesia

  • Analyzed and developed CRM solutions for startups and state-owned companies
  • Led NLP-based Audio Transcriber project for quality assurance using GCP
Nov 2018 — Jul 2021

Vice Head of Programmer / AI Engineer

AMV UI & Tim Robotika — University of Indonesia

  • Managed 6 AI Division members developing autonomous control algorithms for USV and AUV robots
  • Competed in international competitions: AUVSI Roboboat (USA), ASEAN MATE ROV, KKCTBN
  • Built automation and control systems using ROS, Arduino, and computer vision
Education

Academic Background

Master's Degree

Kookmin University

Electrical & Electronics Engineering

Sep 2022 – Aug 2024 · Seoul, South Korea

Bachelor's Degree

University of Indonesia

Computer Engineering · GPA 3.43

Aug 2018 – Sep 2022 · Depok, Indonesia

Stack

Technical Toolbox

Technologies and tools across engineering, research, and development.

Programming Languages

PythonC/C++ JavaC# MATLABJavaScript TypeScriptPHP

AI & Machine Learning

TensorFlowPyTorch LLM Fine-TuningLoRA / QLoRA RAGComputer Vision NLPTensorRT ONNXScikit-learn

Robotics & Hardware

ROS 2NVIDIA Jetson ArduinoLoRa Embedded SystemsSensor Integration

Tools & Platforms

DockerLinux GitGCP AWSFastAPI ReactLaravel PostgreSQL
Contact

Let's build something together.

Available for AI/robotics engineering roles, research collaborations, and consulting projects. Currently based in Seoul, South Korea.

muhammadmiftahfaridh@gmail.com
LinkedIn GitHub Google Scholar ORCID ResearchGate orinu.ai