9+
PROJECTS
2+
YEARS OF EXPERIENCE
3
RESEARCH PAPERS
∞
OPEN TO COLAB
// about me
Who Am I?
I’m an Electrical Engineering and AI/ML engineer building practical AI systems from research to deployment.
My work spans deep learning and computer vision research, LLM-based applications, and IoT systems. I build conversational AI with FastAPI and the OpenAI SDK, run open-source LLMs and vision-language models on local GPU infrastructure, and turn trained models into usable applications and APIs. I also work with ESP32 and embedded systems to connect AI with real-world devices.
My foundation is in deep learning and computer vision, with hands-on experience in PyTorch, data pipelines, model evaluation, and deployment. I am interested in the full AI engineering stack, from model development and prompt engineering to API design, local inference, and edge deployment.
I use autonomous coding agents to move faster, but I do not outsource the engineering thinking. I make the architectural decisions, understand the systems I build, and stay responsible for the implementation from end to end.
CV// what i do
What I Do
Building the future ✦
agents/jeski_agent.py
Memory: 12 turns REASONING
Agentic AI & Multi-Agent Systems
Designing LLM-powered agents and multi-agent workflows with FastAPI and the OpenAI SDK, including tool use, memory, prompt design, and iterative reasoning.
Detection Output
"class": "defect_broken"
"bbox": [142, 88, 37, 41]
Confidence: 98.72%
Computer Vision & Deep Learning
Developing and evaluating object detection models with PyTorch, including YOLO, Faster R-CNN, and RetinaNet, supported by custom datasets and training pipelines.
48GB
VRAM
0
Cloud APIs
24/7
Serving
LLM Deployment & Local Inference
Deploying open-source LLMs and vision-language models on local GPU infrastructure using tools such as Ollama, Open WebUI, and Streamlit.
Sensor Stream
📡 ESP32: telemetry → cloud
DEVICE STATUS
Temp: 26.4°C
Relay: 4/4 OK
WiFi: online
IoT & Embedded Systems
Building ESP32-based systems that integrate sensors, actuators, and control logic for connected devices and practical automation applications.
📊 TRAINING mAP
94%+ mAP
AI Research & Model Evaluation
Conducting research in object detection and data augmentation through controlled experiments, custom dataset construction, and rigorous model evaluation.
// what i built
What I Built
Jeski — Conversational AI Assistant
A context-aware chatbot built with LangChain and the OpenAI SDK, featuring a custom memory system for multi-turn conversations. Prompts were iterated to shape the assistant's behavior, tone, and response quality, then deployed via a live, publicly accessible web interface.
Self-Hosted LocateAnything (VLM)
Deployed NVIDIA's LocateAnything vision-language model on a local RTX 6000 Ada GPU server, with a Streamlit interface for remote, cross-device access over a public network, serving a large AI model with no cloud dependency.
Face Recognition Attendance System
Led AI integration as team leader, deploying and connecting a face-recognition model to cloud infrastructure via IoT and microcontroller hardware. Achieved 98% recognition accuracy under proper lighting with a ~3s average processing time.
Maceration Controller
An ESP32/Arduino controller for maceration processes, coordinating 4-channel relay control, a speed sensor, DS18B20 temperature monitoring, and buzzer alerts for precise, hands-off process control.
// my journey
My Journey
AI Researcher — Coffee Bean Grading System
Syiah Kuala UniversityAug 2025 - July 2026
Developed an end-to-end computer vision pipeline in Python and PyTorch for automated coffee bean defect detection. Built an auto-labeling tool used to annotate 11,000+ images, trained and benchmarked multiple object detection models, and developed a Mosaic Packed Augmentation method for dense-scene detection. Results exceeded 85% mAP on the test set, with 3 research papers authored from the work.
Electrical Engineering Intern
PT PLN (Persero)Jan 2025 - May 2025
Supported the operation and preventive maintenance of diesel power plant equipment, including generator sets, control panels, and auxiliary systems. Monitored electrical and operational parameters, assisted with troubleshooting, and analyzed load and performance data to support system reliability and power distribution.
Vice Leader — KROENG
Syiah Kuala UniversityFeb 2024 - Feb 2025
Coordinated training, project planning, and competition preparation for KROENG, a university community focused on robotics and electrical engineering.
Electronics Laboratory Assistant
Syiah Kuala UniversitySept 2023 - July 2026
Supported 100+ students in electronics, instrumentation, and circuit troubleshooting while reviewing practical work and coordinating laboratory activities.
Robotics & IoT Mentor
Kejar Mimpi AcehOct 2024
Taught IoT fundamentals and mentored 50+ participants in building functional prototypes.
// research
Research
Peer-reviewed work on computer vision, object detection, and automated coffee bean quality grading.
Comparative Study of Object Detection Models for Dense Coffee Bean Defect Detection Using YOLOv11, Faster R-CNN, and RetinaNet
Benchmarks five object detection architectures on dense-scene green coffee bean imagery, comparing accuracy, speed, and robustness for multi-defect detection in real grading conditions.
CoffeeDefect-8: A Multi-Class Dense-Scene Dataset for Green Coffee Bean Defect Detection and Automated Quality Grading
Introduces an 8-class dense-scene dataset of green coffee bean defects with annotation protocol and baselines, enabling reproducible research on automated quality grading.
Advanced Mosaic Augmentation Techniques for Enhancing Multiple Coffee Bean Defect Detection Based on YOLOv11
Explores advanced mosaic augmentation strategies that boost YOLOv11 performance on overlapping, small-object coffee bean defects in dense scenes.
// let's talk
Let's Talk
Have a role, a project, or just a question? My inbox is open — I reply fast.
Let's build something together.
Currently open to projects and collaborations in AI/ML and agentic systems.
Email me here