About
I spent over two years at a Taiwanese-owned Adidas manufacturer in Indonesia — redesigning sewing workflows, managing cross-functional teams, and leading a legacy-to-SAP data migration. That work taught me how decisions stall when data is scattered and how much clarity a single clean dataset can unlock.
Having completed a Master of Applied Data Analytics at the Australian National University (GPA 6.5 / 7.0), I apply that operational mindset to NLP pipelines, LLM evaluation, and statistical modelling. My recent research at ANU involved building a bilingual NER pipeline processing ~7,900 news articles and designing an inter-model agreement framework across four frontier LLMs.
Python · R · SQL / PostgreSQL
Google ADK · Gemini · LangChain · spaCy · PyTorch · scikit-learn · sentence-transformers · FAISS · Ollama
LLM agents & tool use · tiered memory systems · semantic retrieval · Bayesian modelling · regression · time series analysis · NER · LLM evaluation · network analysis
English (C1) · Indonesian (Native) · Korean (Conversational)
Projects
A from-scratch implementation of the MemGPT paper's tiered-memory architecture, giving an LLM agent persistent memory across sessions. Six tools operate over three storage tiers — always-in-prompt core memory, keyword-searchable recall (SQLite FTS5), and semantically-searchable archival (FAISS with 384-dim sentence-transformer embeddings) — with human-in-the-loop gating enforcing strict-undo semantics on the one destructive tool. Built on Google ADK with Gemini, deployed as a Gradio interface on Hugging Face Spaces.
A time series study of bituminous coal production built in R Markdown, examining how output has changed over time and the trend and seasonal patterns that shape it — published as a reproducible report on Posit Connect Cloud.
A PyMC model that uncovers latent household shopping segments from the Dunnhumby "Complete Journey" grocery transaction dataset. Incorporates product-level covariates — price, discounts, and popularity — to distinguish consumer-driven from retailer-influenced purchase behaviour through a hierarchical Bayesian mixture framework.
A three-stage data mining pipeline — k-means clustering, decision tree classification, and Apriori association rule mining — applied to the 2025 AES Aboriginal and Torres Strait Islander Voter Extension Study (n=802). Uncovers distinct voter engagement profiles, identifies institutional trust and Indigenous policy salience as the strongest predictors of vote choice, and surfaces attitude bundles that co-occur with Labor, Coalition, and Greens support.
A multinomial logistic regression model estimating the probability of winning, drawing, or losing a chess game from Chess.com data scraped across four players (2024–2026). Tests ELO rating difference, piece-specific move counts, colour, and time control as predictors, selects the best-fitting specification via AIC/BIC, and applies it to predict outcomes against world champion Ju Wenjun.
A pipeline for merging organisation mentions that appear under different names — Indonesian versus English variants, acronyms, formatting differences — into canonical forms for accurate frequency counting and network analysis. Uses Ollama as the LLM router to intelligently group bilingual organisation name variants.
An exploratory data analysis using the Pokémon dataset from the pokemon R library, modelling how speed points vary by primary type, secondary type, and weight through regression analysis and visualisation.
Experience
AI Engineer Internship
VentureSEA Digital Singapore
Jul 2025 — Present
Built a chat agent with three kinds of memory — sticky notes always in view (in-prompt), a searchable log of past messages (SQLite + FTS5), and a notebook for long-form notes retrieved by meaning (FAISS + sentence-transformers) — all wired together on Google Agent Development Kit and deployed as a Gradio demo on Hugging Face Spaces.
Applied NLP & AI Internship in Political Science
Australian National University
Dec 2025 — Present
Building NLP infrastructure for a climate advocacy research project — bilingual NER pipeline, network graph construction, and survey design for Indonesian NGO mapping.
Applied AI Internship in Political Science
HASS Digital Research Hub in Australian National University
Jul 2025 — Oct 2025
Designed an inter-model agreement framework for validating LLM-powered political text analysis, benchmarking four frontier models and iterating prompt quality.
Costing Supervisor
Adidas ODM (Shyang Shin Bao Industrial)
May 2024 — Dec 2024
Built centralised cost frameworks, validated data during a legacy-to-SAP migration, and produced executive-facing margin reports across product development.
Technical Project Manager
Adidas ODM (Shyang Shin Bao Industrial)
Sep 2022 — May 2024
Redesigned sewing workflows achieving a 4× cycle-time reduction across 250K pairs, managed a 5-person team, and sustained 651K pairs monthly output.
Research Assistant
Korea University
2021 — 2022
Configured automated wastewater simulations and processed time-series data for an antibiotic resistance monitoring project.
Education
Master of Applied Data Analytics
The Australian National University
2025 — Jul 2026
GPA: 6.5 / 7.0
Key courses: Data Mining, Statistical Modelling, Generalised Linear Modelling, Relational Databases (PostgreSQL), Data Wrangling, Document Analysis.
Bachelor of Chemical Engineering
Universitas Surabaya
2016 — 2020
GPA: 3.897 / 4.000