Hello

I am Yavie Azka.

AboutMe

Undergraduate Informatics student at Institut Teknologi Bandung and Tanoto Foundation Scholar, focused on machine learning and applied AI, from data preprocessing and statistical analysis to model development and production deployment. Currently building production ML systems for a PhD clinical research project on liver cirrhosis outcome prediction.

Institut Teknologi Bandung · GPA: 3.77/4.00Tanoto Foundation ScholarAI/ML & Software Engineer

WorkExperience

AI & Software Engineer

Production-Deployed
2026 — Present
ChirroSmartAI—Clinical Decision Support System (CDSS) for Hepatologycirrhosmart.site
Publication: Co-Author, Computer Methods and Programs in Biomedicine (CMPB) — Q1 Scopus (Under Review)

Built the end-to-end ML pipeline for a PhD clinical research project on liver cirrhosis outcome prediction, collaborating within a 6-person team of physician specialists, pharmacology researchers, and PhD candidates.

  • Cleaned and imputed messy real-world clinical data using MICE via Random Forest and majority-vote logic, resolving outliers and missing values.
  • Evaluated predictive signals via statistical analysis (Point-Biserial, Cramér's V, Mutual Information), uncovering a clinical "survival bias" in Length of Stay outcomes.
  • Built and validated classification models (Random Forest, XGBoost) with Stratified 5-Fold Cross-Validation, achieving AUROC 0.84 and F1-score 0.60 for mortality prediction (Random Forest, n=395), addressing class imbalance via threshold analysis and MCC evaluation.
  • Built a rule-based engine for hepatotoxic drug dose recommendation and adjustment, covering 8 critical drug combinations.
  • Deployed models to production via Vercel Serverless Functions, integrated with a Next.js and Prisma ORM frontend for real-time clinical inference.
Machine LearningPythonscikit-learnXGBoostpandasNumPyNext.jsReactPrisma ORMVercel Serverless

FeaturedProjects

content-detector.ts
MATCH: [SLOT88 / G4C0R]Levenshtein Dist: 0.2 (Fuzzy Homoglyph)Aho-Corasick: Multi-Pattern Trie HitTesseract.jsWASM OCRDOM TreeWalker Scanning Active · Offline
AI & Pattern Security

Judol Detector — Browser Content Moderation

Automated Chromium extension for real-time detection and masking of online gambling promotion using multi-pattern string algorithms and offline computer vision OCR.

  • Exact matching via KMP, Boyer-Moore, & Aho-Corasick trie automaton (O(n + z))
  • Fuzzy visual manipulation & leetspeak detection using Weighted Levenshtein Distance
  • Offline image text extraction powered by local Tesseract.js WebAssembly
TypeScriptManifest V3Aho-CorasickTesseract.js OCRLevenshtein
compiler-pipeline.cpp
DFA LexerDecorated ASTStack VM001: LIT 0 42002: STO 0 3; assign003: CAL 1 12004: OPR 0 13; call / eval005: RET 0 0006: JMP 0 0; execute OK
Systems & Compilers

Arion Compiler & Virtual Machine

Complete language compilation pipeline from source code to bytecode execution on a custom stack-machine interpreter built from scratch in C++17.

  • DFA-based Lexer & Recursive Descent Parser producing Decorated Abstract Syntax Trees
  • Type system checking, nested scope resolution, and symbol table management
  • Three-Address Code (TAC) intermediate generation executing on a Stack Machine VM
C++17Compiler DesignDFA LexerDecorated ASTStack Machine VM
osi-stack.cpp
L7: HTTP Web Server & DHCP ServerPORT 80/67L4: TCP [SYN → SYN-ACK → ACK] State MachineESTABLISHEDL3: IPv4 Routing & ICMP Echo RequestROUTINGL2: Ethernet Frame & ARP Table CacheMAC LAYER
Distributed Systems

Magi System — OSI Network Protocol Simulator

Comprehensive network protocol simulator written in pure C++ without OS network socket libraries, modeling packet lifecycles across OSI Layer 2 to Layer 7.

  • Custom Ethernet frames, ARP table caching, and multi-subnet IPv4 routing logic
  • Full TCP state machine: 3-way handshake, sequence numbers, and connection teardown
  • Application layer implementation including dynamic DHCP server/discover and HTTP server
C++17NetworkingTCP State MachineOSI StackCMake
tsp-heuristic.py
v0T1v1v2T2v3v4O(V² log V) Heuristic
Algorithm Engineering

HUTAO-TSP — Teleportation Graph Heuristic

Novel heuristic algorithm approximating optimal TSP tours on weighted graphs featuring non-Euclidean teleportation portals in polynomial time.

  • Combines Minimum Spanning Tree (MST) routing with custom branch evaluation DFS
  • Guarantees polynomial-time complexity while outperforming standard greedy baselines
  • Benchmarked on synthetic and geometric graph topologies with NumPy
PythonGraph TheoryHeuristicsNumPyAlgorithm Design
octree-sat.cpp
SAT Axis TestOpenMP: 8 ThrdOBJ → Voxel
High Performance Computing

3D Voxelizer — Octree & Parallel SAT

High-throughput 3D mesh voxelization engine converting OBJ geometries into discrete spatial voxel trees using recursive spatial subdivision and OpenMP concurrency.

  • Divide-and-conquer Octree spatial tree with Separating Axis Theorem (SAT) intersection
  • Multi-threaded parallel acceleration with OpenMP for deep-level spatial decomposition
  • Dynamic empty space pruning minimizing RAM allocation and traversal overhead
C++OpenMPOctreeComputer GraphicsParallel Computing
dom-traversal.tsx
<html><body><head><main><nav><title><meta>Query: main > div | BFS: 1.4ms | LCA Binary Lifting
Full-Stack & Algorithms

DOM PathFinder & Tree Visualizer

Web-based DOM tree scraper and interactive graph visualizer exploring HTML hierarchies via Breadth-First Search (BFS) and Depth-First Search (DFS) with CSS Selectors.

  • Interactive node traversal animation with Lowest Common Ancestor (LCA) Binary Lifting
  • Full-stack Next.js 15 App Router architecture containerized with Docker
  • Deployed to Microsoft Azure Cloud App Service with production CI/CD
Next.js 15TypeScriptTree Algorithms (BFS/DFS)DockerAzure

TechnicalSkills

Comprehensive technical toolkit across applied AI, systems programming, and modern full-stack development.

PythonCore

Primary ML Language

scikit-learnCore

Pipelines & Cross-Val

XGBoostCore

Ensemble Modeling

pandasCore

Clinical Data Processing

NumPyCore

Numerical Computing

Statistical Analysis

Cramér's V, Point-Biserial

MICE Imputation

Missing Value Resolution

Model Evaluation

AUROC, F1, MCC Analysis

C / C++Core

Compilers & CP

TypeScriptCore

Type-Safe Systems

Java

OOP & JavaFX

SQL

Relational Queries

JavaScript

Modern ES6+

Prolog

Logic Programming

Next.jsCore

App Router & SSR

ReactCore

Modern Component UI

Tailwind CSSCore

Responsive Theming

Prisma ORM

Database Modeling

PostgreSQL

Schema & Storage

Vercel Serverless

Real-time Inference

JavaFX

Desktop Interfaces

Git & GitHubCore

Version Control & CI

Linux / CLI

Shell & Workflows

Compiler Construction

Lexer & Stack VM

Competitive ProgrammingCore

National Olympiad Level

Languages & Communication

Indonesian(Native)

First language

English(Professional Working Proficiency)

1st Winner Provincial English Debate Competition (SMAN 3 Bukittinggi)

Achievements

2024

  • 2nd Winner - National Informatics Olympiad

    Tensai 13th SMAN 10 Padang

  • 2nd Winner - Provincial Informatics Olympiad

    Festival Smantripa IX SMAN 3 Painan

  • 3rd Winner - Provincial Informatics Olympiad

    Fesmative 2nd SMAN 4 Padang

  • 1st Winner - Provincial English Debate Competition

    SMAN 3 Bukittinggi

2023

  • 1st Winner - National Mathematics Challenge

    PSB Universitas Andalas

  • 1st Winner - Provincial Informatics Olympiad

    Technofest Universitas Andalas

  • 1st Winner - Provincial Informatics Olympiad

    Tensai 12th SMAN 10 Padang

  • 4th Winner - Sumatra Regional Olympiad

    SMAPSIC XVIII SMAN 1 Padang

  • 2nd Winner - Provincial Informatics Olympiad

    SMAN 1 Gunung Talang

2022

  • 2nd Winner - National Informatics Olympiad

    Lustrum Universitas Andalas

  • 2nd Winner - Sumatra Regional Olympiad

    SMAPSIC XVII SMAN 1 Padang

  • 3rd Winner - Provincial Informatics Olympiad

    Tensai 11th SMAN 10 Padang

  • 2nd Winner - Provincial English Debate Competition

    SMAK Kalam Kudus Padang