ML Engineer | Quant Developer | Product Engineer
I build production-grade ML systems, quantitative analytics platforms, and end-to-end products that ship to users.
- ML/AI Systems: CNN-based computer vision models with >90% accuracy, deployed via ONNX/TFLite for mobile inference
- Quantitative Analytics: IPO underwriter performance platforms with automated ETL, real-time dashboards, and multi-stock forecasting engines
- Full-Stack Products: Mobile apps published to Play Store, web platforms with searchable databases, and automated data pipelines
Problem: Manual plastic sorting is slow and error-prone — mislabeled resin types contaminate recycling streams, and most
people don't know how to dispose of HDPE, PET, PP, or PS correctly.
Achievement: Built an end-to-end plastic classification product: photograph or upload an image, get the resin type (HDPE,
PET, PP, PS) with confidence scoring, and read Bahasa Indonesia recycling guidance — ResNet18 fine-tuned in PyTorch, Flask REST
API, Next.js frontend with browser camera support; Highly Commended / Top Finalist at AI for Sustainability Hackathon.
Stack: PyTorch, ResNet18, Flask, Next.js, TypeScript, Tailwind CSS
Status: Live at Google Cloud Run ·
GitHub
Problem: Indonesian traders piece together quotes, charts, technical signals, and market context across disconnected
tools — no single bilingual flow from ticker search to actionable analysis.
Achievement: Built a production equities terminal with live IDX charts, RSI/MACD/Bollinger signals, Pulse AI summaries
(OpenAI/DeepSeek), watchlists, screeners, and automated ETL — evolved from a single IPO analytics app shipped to Google into an end to end tools for complex AI and Agent Orchestration Analysis .
Stack: Next.js 16, React 19, TypeScript, Supabase, Bun, OpenAI, DeepSeek, Vercel
Status: Live at ipo-tracker.site · Google
Play Store
Problem: University ML and computer vision theory — decision boundaries, SVM margins, clustering, PCA, convolution,
camera models — is hard to internalize from static slides and exam notes alone.
Achievement: Built a two-part interactive study series from BINUS coursework: Interactive ML (Sessions 15–22) with
live D3 visualizations for decision trees, SVM, K-Means/DBSCAN, and PCA/LDA/t-SNE; Interactive CV with hands-on labs for
spatial convolution, feature detection, pinhole projection, and epipolar geometry — upload images, edit kernels, and
explore 3D modules with KaTeX math.
Stack: React 19, TypeScript, Vite, D3, Next.js 16, React Three Fiber, Three.js, Framer Motion, KaTeX
Status: Live at interactive-ml-drab.vercel.app ·
interactive-cv-psi.vercel.app · ML
repo · CV repo
Problem: Apple Health data export is XML-based and difficult to analyze programmatically.
Achievement: Built Swift parser converting XML to structured JSON with aggregation, summaries, and export features.
Stack: Swift, XML parsing, data transformation
Problem: Optimizing nutrition, budget, and food choices simultaneously is difficult.
Achievement: Built a nutrition-aware grocery optimization platform with structured data, pricing logic, and recommendation-driven UX.
Stack: React, TypeScript, Vite, Tailwind CSS
ML/AI
PyTorch, XGBoost, TensorFlow Lite, ONNX, scikit-learn, CNNs, time-series forecasting
Backend
Python (Django, FastAPI, Flask), Node.js, Express, ASP.NET Core MVC, Entity Framework Core, PostgreSQL
Frontend & Mobile
React, React Native, Next.js, Flutter, Kivy, Android Studio
Data / Automation
pandas, web scraping, Make.com, n8n, ETL pipelines, data normalization
Systems
C/C++ (data structures, systems programming), Java, Swift, performance optimization
- AI for Sustainability Hackathon: Highly Commended / Top Finalist (PlastiTrace) — >99% CNN accuracy
- FinTech Hackathon: Shipped UW-Tracker Mobile to Google Play Store within 48 hours
- Published Apps: Google Play Store publication (UW-Tracker Mobile)
- Research Output: Authored LaTeX scientific reports with PGFPlots, ablation studies, and deployment handbooks
- Production Systems: Built 10+ production-level tools across ML, quant finance, and mobile development
- Team Leadership: Led multiple hackathon teams with end-to-end delivery from research to deployment