Benedict Pepper | CS Portfolio

About Me

Driven by Curiosity, Defined by Code.

I am Benedict Pepper, a computer science student at Ma Chung University. I build machine learning systems and the software around them, from the model backend to the web interface. I want to make complex technical code accessible through tools that are simple to use.

My projects usually focus on medical imaging and data analysis. I recently finished a research internship where I optimized Bitcoin trading strategies, using Numba to run grid searches on 14 years of market data. Before that, I worked on deep learning models that segment fetal ultrasound scans on low cost hardware. I also write about sign language recognition and chest X-ray classification. I enjoy writing clean APIs, designing database schemas, and building interfaces that work.

STACK

Python

Java

JavaScript

TypeScript

Lua

PyTorch

TensorFlow

XGBoost

Numba

Next.js

React

FastAPI

MongoDB

PostgreSQL

Docker

ONNX Runtime

Gemini API

Experiences

A timeline of my academic, professional, and leadership growth.

Technical Researcher (PKM-KC)

National Student Creativity Program | March 2026 - June 2026

Collaborated on FitMate, a medication safety scanner for Traditional Chinese Medicine (TCM). Built a working PWA prototype that translates TCM labels via Gemini multimodal OCR, cross-references a toxicity database, and integrates a WhatsApp chatbot for guidance.

Quantitative Research Intern (PKL) photo 1

Quantitative Research Intern (PKL)

Ma Chung University | Feb 2026 - July 2026

Built a Numba JIT-compiled grid-search engine optimizing Bitcoin trading thresholds and asset allocation across 1.29 million parameter combinations over 14 years of market data. Identified Logarithmic Regression as the strongest on-chain signal and established a strict In-Sample and Out-of-Sample validation framework to prevent overfitting.

Programming & Coding Instructor

Kalam Kudus & Kawai Edulab | Feb 2025 - Present

Teach programming (Python, Lua in Roblox Studio, and Scratch) to 80+ elementary and junior high students weekly across two institutions. Develop curriculum covering core data structures, algorithms, and game development.

Treasurer & Event Manager

Informatics Engineering Student Association (HMP) | Oct 2023 - Jun 2025

Managed financial planning and cross-functional teams for major university programs. Coordinated the 'IT School' workshop for 100+ high school students and led the 'Harmony IT' outdoor integration program, successfully optimizing the budget by 53%.

Project Manager & Lead Dev

Academic Projects | 2023 - 2024

Spearheaded full-stack and data science projects, coordinating with team members to deliver high-quality applications like 'Lapor Aman' and various analytical tools under tight deadlines.

Community & Event Volunteer

Various Initiatives | 2023 - Present

Actively engaged in diverse community initiatives. Served on the Health Division for the Ma Chung Festival committee and managed operations for the Smile 3.0 blood donation event. Handled social media and e-commerce platforms for Riverkids Special Education School. Additionally, contributed to campus sports events like the Malang Sportival run and conducted technical workshops for high school students.

Selected Works

FitMate: TCM Safety Scanner screenshot 1

FitMate: TCM Safety Scanner

A Progressive Web App that helps consumers safely navigate Traditional Chinese Medicine products. Users scan TCM labels with their phone camera — the app extracts and translates Mandarin ingredient text via a Gemini multimodal pipeline, then cross-references a validated toxicity database to flag dangerous compounds and contraindications. When warnings are detected, users are bridged to a rule-based WhatsApp chatbot for hallucination-free medical guidance. Built with a Next.js PWA frontend, Python FastAPI backend, and MongoDB knowledge base.

Next.jsFastAPIMongoDBGemini APIWhatsApp APIPWA

View Repo Live Demo

Fetal Head Ultrasound Segmentation

A deep learning project comparing lightweight encoders (ResNet-50, EfficientNet-B0, ConvNeXt V2 Atto) for fetal head circumference segmentation under simulated low cost ultrasound hardware noise. Designed for edge deployment, the best model runs efficiently on low memory devices.

PyTorchONNX RuntimeDockerPython

View Repo

Brain Tumor FCM Segmentation screenshot 1

Brain Tumor FCM Segmentation

An automated medical image analysis tool that applies the Fuzzy C-Means (FCM) soft clustering algorithm to segment brain tumor regions from MRI scans. Built as a Streamlit web application, users can adjust the number of clusters and fuzziness parameter to analyze segmentation results in real time.

PythonStreamlitFCMMedical Imaging

View Repo Live Demo

BTC Strategy Lab

An interactive Bitcoin backtesting simulator. Users run backtests with customizable threshold and asset allocation parameters across 14 years of BTC data, powered by a Numba JIT-compiled engine. The dashboard displays results from 1.29 million combinations with equity curves, signal overlays, and sensitivity heatmaps.

PythonStreamlitNumbaBacktestingOn-chain Analytics

View Repo Live Demo

WorldTour AR: Landmarks Exploration screenshot 1

WorldTour AR: Landmarks Exploration screenshot 2

WorldTour AR: Landmarks Exploration screenshot 3

WorldTour AR: Landmarks Exploration

An interactive mobile Augmented Reality application developed in Unity with Vuforia Engine and Universal Render Pipeline. It projects low-poly 3D models of historical landmarks into the real world, featuring educational trivia quizzes and local progress tracking.

UnityVuforia EngineURPC#

View Repo

FFmpeg Web UI: Visual Converter

A graphical interface for FFmpeg that allows users to convert audio and video files via a web browser. It simplifies the process of adjusting bitrates and formats through a drag-and-drop system. The app also displays the exact FFmpeg command being executed in the background, serving as both a utility and a reference for learning terminal commands.

Python FlaskFFmpegJavaScript

View Repo Live Demo

Academic Research

Optimalisasi Parameter Trading Bitcoin Menggunakan Grid Search pada Tiga Metrik Evaluasi Berbasis Indikator Logarithmic Regression

Completed PKL Research (2026)

Keywords: Quantitative Finance, Optimization, Backtesting, On-chain Metrics

This research develops a quantitative optimization framework to find the ideal combination of trading parameters based on historical data. It optimizes buying and selling thresholds and asset allocation using the CBBI indicator to maximize Total Return and Sharpe Ratio while minimizing Maximum Drawdown. The study validates performance through In-Sample and Out-of-Sample testing and involves an interactive simulation dashboard.

View Research Repo

Efficiency-Robustness Trade-off of Fetal Head Segmentation Architectures for Resource-Constrained Ultrasound Deployment

Insyst Journal (First Author)

Keywords: Deep Learning, PyTorch, Edge Deployment, Medical Ultrasound, ONNX Runtime

A research paper evaluating lightweight deep learning encoders (ResNet-50, EfficientNet-B0, ConvNeXt V2 Atto) for segmenting fetal head circumference in low cost ultrasound hardware. The best model achieved 0.977 Dice accuracy at 322ms inference speed, optimized for edge execution.

Download Manuscript View Research Repo

FitMate: An Innovative Mandarin Herbal Medicine Label Scanner

Completed PKM-KC Research (2026)

Keywords: Health-Tech, Chatbot, Python, WhatsApp API, Patient Safety

An interdisciplinary research project funded under the national Student Creativity Program (PKM-Karsa Cipta). FitMate is an innovative scanner and translation tool for Mandarin herbal medicine (TCM) labels. Accompanied by a website database and WhatsApp chatbot, it aims to prevent medication errors, educate users on side effects and drug interactions, and help screen products containing undisclosed medicinal chemicals (BKO) to improve patient safety.

Download Whitepaper

Studi Komparatif XGBoost dan MLP Berdasarkan Evaluasi Fitur pada Klasifikasi Angka Bahasa Isyarat SIBI

Ma Chung Proceedings (First Author)

Keywords: Applied Intelligence Systems, Python, XGBoost, MLP, Sensor Data

A research project aimed at improving the classification accuracy of sign language numbers using data collected from a sensor-equipped glove. Analyzing a dataset of 45,000 samples across 20 distinct classes and 16 physical sensor features, the study implements XGBoost and Multi-Layer Perceptron (MLP) networks to exceed the baseline 98.9% kNN accuracy and resolve overlapping classes.

Download Manuscript View Research Repo

Pengaruh Progressive Resizing pada EfficientNetB4 untuk Klasifikasi Pneumonia Bakterial dan Viral

Ma Chung Proceedings (First Author)

Keywords: Deep Learning, EfficientNet, Transfer Learning, Progressive Resizing

A comparative study on training strategies for classifying Bacterial vs. Viral Pneumonia using Pediatric Chest X-Rays. The research demonstrated that fixed-size resolution training outperformed progressive resizing, achieving a final accuracy of 87.14% and an ROC AUC of 0.9431, optimizing computer-aided diagnostic reliability.

Download Manuscript View Research Repo

Revitalizing the River Kids SLB Autism E-Commerce Website

Academic Manuscript

Keywords: UI/UX Design, System Usability Scale (SUS), Web Accessibility

Conducted a participatory UI/UX redesign of an e-commerce platform for special needs students' crafts. Through quantitative evaluation using the System Usability Scale, the intervention yielded a 25.4% improvement (from 57.81 to 72.50), elevating the platform's category from 'Marginal' to 'Good' and improving public visibility.

Download Manuscript

Survey of Language Style Usage in Informatics Engineering Slides

Course Research Report

Benedict
Pepper.

About Me

Driven by Curiosity, Defined by Code.

Experiences

Technical Researcher (PKM-KC)

Quantitative Research Intern (PKL)

Programming & Coding Instructor

Treasurer & Event Manager

Project Manager & Lead Dev

Community & Event Volunteer

Selected Works

FitMate: TCM Safety Scanner

Fetal Head Ultrasound Segmentation

Brain Tumor FCM Segmentation

BTC Strategy Lab

WorldTour AR: Landmarks Exploration

FFmpeg Web UI: Visual Converter

Academic Research

Optimalisasi Parameter Trading Bitcoin Menggunakan Grid Search pada Tiga Metrik Evaluasi Berbasis Indikator Logarithmic Regression

Efficiency-Robustness Trade-off of Fetal Head Segmentation Architectures for Resource-Constrained Ultrasound Deployment

FitMate: An Innovative Mandarin Herbal Medicine Label Scanner

Studi Komparatif XGBoost dan MLP Berdasarkan Evaluasi Fitur pada Klasifikasi Angka Bahasa Isyarat SIBI

Pengaruh Progressive Resizing pada EfficientNetB4 untuk Klasifikasi Pneumonia Bakterial dan Viral

Revitalizing the River Kids SLB Autism E-Commerce Website

Survey of Language Style Usage in Informatics Engineering Slides

Let's Connect

Send a Message

BenedictPepper.

About Me

Driven by Curiosity, Defined by Code.

Experiences

Technical Researcher (PKM-KC)

Quantitative Research Intern (PKL)

Programming & Coding Instructor

Treasurer & Event Manager

Project Manager & Lead Dev

Community & Event Volunteer

Selected Works

FitMate: TCM Safety Scanner

Fetal Head Ultrasound Segmentation

Brain Tumor FCM Segmentation

BTC Strategy Lab

WorldTour AR: Landmarks Exploration

FFmpeg Web UI: Visual Converter

Academic Research

Optimalisasi Parameter Trading Bitcoin Menggunakan Grid Search pada Tiga Metrik Evaluasi Berbasis Indikator Logarithmic Regression

Efficiency-Robustness Trade-off of Fetal Head Segmentation Architectures for Resource-Constrained Ultrasound Deployment

FitMate: An Innovative Mandarin Herbal Medicine Label Scanner

Studi Komparatif XGBoost dan MLP Berdasarkan Evaluasi Fitur pada Klasifikasi Angka Bahasa Isyarat SIBI

Pengaruh Progressive Resizing pada EfficientNetB4 untuk Klasifikasi Pneumonia Bakterial dan Viral

Revitalizing the River Kids SLB Autism E-Commerce Website

Survey of Language Style Usage in Informatics Engineering Slides

Let's Connect

Send a Message

Benedict
Pepper.