Nathaniel Jonathan Rusli
AI and Data Enthusiast
Click Me to Chat with NateBot!
Hey, I'm Nathaniel! 👋🏻 I'm heading into my final year of Computer Science at ITB.
I'm a massive nerd when it comes to AI, Data, and Finance—and I've snagged a few national wins in data science and equity research along the way!
Try chatting with NateBot and keep scrolling to check out my achievements and projects I've built!
Projects
AI-Powered Cinematic Portfolio & RAG Engine (This Website)
Architected a high-fidelity portfolio featuring an embedded Retrieval-Augmented Generation (RAG) chatbot. Engineered an advanced retrieval strategy utilizing query decomposition and multi-query routing via LangChain to handle complex user interactions. Orchestrated the entire deployment pipeline using Docker and GitHub Actions for a robust, containerized delivery on AWS infrastructure.
Automated Stock Tearsheet Generator with Batch Pipeline
Architected an end-to-end financial Data Lakehouse implementing a Medallion architecture on Amazon S3. Engineered a robust ETL pipeline using Apache Airflow and PySpark for large-scale batch processing, integrated with an LLM engine to extract quantitative sentiment from unstructured market news. Designed a decoupled, serverless presentation layer utilizing AWS Athena, AWS Glue and FastAPI to serve interactive tearsheets and technical charts via plotly.
Pothole Semantic Segmentation with Statistical EDA and EoMT
Architected a high-precision semantic segmentation pipeline to automate the detection of road wear and potholes from drone imagery. The methodology involved rigorous statistical EDA to handle class imbalances before training. Deployed an Evaluation of Model Trust (EoMT) framework alongside PyTorch to ensure the reliability of predictions. Won 1st Place at the National Informatics Festival Unpad for this research.
Cultural Image Classification with DINOv3 and Explainable AI (XAI)
Architected a state-of-the-art cultural image classifier using the DINOv3-Huge vision transformer with a custom classification head for fine-grained Indonesian cultural heritage recognition. Conducted extensive transfer learning from self-supervised pretraining, integrating feature extraction, hyperparameter optimization, and explainable AI techniques to interpret model decisions on complex visual patterns. Delivered robust performance through meticulous fine-tuning and comprehensive evaluation, securing 5th place nationally in the Logika UI Data Science Competition.
Mealdrop
Engineered scalable backend microservices and responsive mobile interfaces as a Software Engineer for the Mealdrop platform. Designed and optimized complex relational database schemas in PostgreSQL to handle high-volume, concurrent transaction data efficiently. Streamlined the server infrastructure on AWS, ensuring high availability and seamless data synchronization with the Flutter client.
AI-Powered Cinematic Portfolio & RAG Engine (This Website)
Architected a high-fidelity portfolio featuring an embedded Retrieval-Augmented Generation (RAG) chatbot. Engineered an advanced retrieval strategy utilizing query decomposition and multi-query routing via LangChain to handle complex user interactions. Orchestrated the entire deployment pipeline using Docker and GitHub Actions for a robust, containerized delivery on AWS infrastructure.
Cultural Image Classification with DINOv3 and Explainable AI (XAI)
Architected a state-of-the-art cultural image classifier using the DINOv3-Huge vision transformer with a custom classification head for fine-grained Indonesian cultural heritage recognition. Conducted extensive transfer learning from self-supervised pretraining, integrating feature extraction, hyperparameter optimization, and explainable AI techniques to interpret model decisions on complex visual patterns. Delivered robust performance through meticulous fine-tuning and comprehensive evaluation, securing 5th place nationally in the Logika UI Data Science Competition.
Binomial Option Pricing Model (BOPM) with Dynamic Programming (DP)
Engineered a computational finance engine to calculate the theoretical fair value of derivatives using the Binomial Option Pricing Model. Designed a highly optimized, lattice-based algorithm leveraging Dynamic Programming (DP) to efficiently compute early exercise boundaries and terminal payoffs. This approach drastically reduced the time complexity for deep, multi-period market pricing simulations.
Automated Stock Tearsheet Generator with Batch Pipeline
Architected an end-to-end financial Data Lakehouse implementing a Medallion architecture on Amazon S3. Engineered a robust ETL pipeline using Apache Airflow and PySpark for large-scale batch processing, integrated with an LLM engine to extract quantitative sentiment from unstructured market news. Designed a decoupled, serverless presentation layer utilizing AWS Athena, AWS Glue and FastAPI to serve interactive tearsheets and technical charts via plotly.
Mealdrop
Engineered scalable backend microservices and responsive mobile interfaces as a Software Engineer for the Mealdrop platform. Designed and optimized complex relational database schemas in PostgreSQL to handle high-volume, concurrent transaction data efficiently. Streamlined the server infrastructure on AWS, ensuring high availability and seamless data synchronization with the Flutter client.
Mini x86 Operating System
Architected a custom 32-bit x86 operating system from the ground up to establish fundamental low-level bare-metal execution. Engineered intricate memory management systems utilizing explicit hardware paging and a Global Descriptor Table for secure memory isolation. Designed a preemptive process scheduler and integrated an ext2-compatible file system to orchestrate concurrent multitasking and resilient data persistence.
Pothole Semantic Segmentation with Statistical EDA and EoMT
Architected a high-precision semantic segmentation pipeline to automate the detection of road wear and potholes from drone imagery. The methodology involved rigorous statistical EDA to handle class imbalances before training. Deployed an Evaluation of Model Trust (EoMT) framework alongside PyTorch to ensure the reliability of predictions. Won 1st Place at the National Informatics Festival Unpad for this research.
E-commerce with Real Time Auction
Architected a high-concurrency e-commerce and real-time auction engine leveraging ACID-compliant database locking mechanisms to ensure deterministic bid resolution and prevent race conditions during high-frequency trading. Engineered bidirectional, event-driven WebSocket channels to instantly propagate market state mutations, bid histories, and precise countdown synchronizations across distributed client architectures. Designed a robust cursor-based data ingestion pipeline and an asynchronous push-notification cryptosystem using VAPID authentication, scaling efficiently to resolve concurrent connections while maintaining strict transactional data integrity.
Sobel Detection Parallel Programming
Engineered a high-performance image processing engine implementing the Sobel operator using CUDA for SIMT and GPU acceleration. This project demonstrates extreme proficiency in high-performance computing (HPC) and low-level hardware optimization for latency-critical tasks.
Achievements
- Most Outstanding Student of Global Consumer Intelligence (GCI) UTokyo09
- 2nd Winner of ARA Data Science Competition08
- 1st Winner of IFEST Data Analytics Competition07
- 5th Place Finalist of Logika UI Data Science Competition06
- Top 5 Finalist of INFEST Equity Research Competition05
- 1st Winner of Talent Growth x MarkPlus Business Case Competition04
- 1st Winner of StudentxCEOs Business Case Competition03
- 3rd Winner of Investing Analysis Competition02
- Bakti BCA Scholarship Awardee01
Tap anywhere to close