rss_feed
01.About MeTentang Saya
Saya adalah seorang AI & Software Engineer yang mendedikasikan diri untuk menjembatani riset akademis dengan implementasi sistem skala produksi. Perjalanan saya berawal dari Pendidikan Fisika — tempat saya membangun dasar pemikiran berbasis prinsip pertama (first-principles thinking) dan pedagogi — hingga menempuh Magister Teknik Informatika untuk mendalami rekayasa kecerdasan buatan. Melalui perpaduan unik ini, saya merancang dan membangun sistem interaktif serta pipeline AI yang aman, berkinerja tinggi, dan siap pakai.I am an AI & Software Engineer dedicated to bridging academic research with production-grade system implementations. My journey began in Physics Education — where I developed a foundation in first-principles thinking and pedagogy — and evolved into a Master's in Informatics Engineering to specialize in artificial intelligence. Through this unique combination, I design and build interactive applications and secure, high-performance AI pipelines.
Berfokus pada pengembangan sistem Generative AI dan MLOps. Saya memiliki keahlian dalam merancang agen LLM kustom, arsitektur RAG hibrida menggunakan pgvector dan Elasticsearch, mitigasi halusinasi model, serta instruction fine-tuning (LoRA). Untuk menjamin keandalan produksi, saya membangun layanan inferensi cepat menggunakan FastAPI (Python), orkestrasi kontainer (Docker & Kubernetes), serta menerapkan proxy keamanan input/output menggunakan guardrails untuk sistem enterprise.Focused on Generative AI and MLOps systems. I specialize in building custom LLM agents, hybrid RAG architectures using pgvector and Elasticsearch, model hallucination mitigation, and instruction fine-tuning (LoRA). To ensure production reliability, I engineer high-performance inference services using FastAPI (Python), container orchestration (Docker & Kubernetes), and deploy input/output security proxies using guardrails for enterprise systems.
Core FocusFokus Utama
smart_toy
Generative AI & RAG
Custom LLM agents, RAG architectures with pgvector & Elasticsearch, and security guardrails.Agen LLM kustom, arsitektur RAG dengan pgvector & Elasticsearch, dan guardrails keamanan.
2 deployments · 1 Scopussettings_suggest
MLOps & Pipelines
LoRA instruction-tuning, experiment tracking with MLflow, model versioning with DVC, and FastAPI.Instruction-tuning LoRA, pelacakan eksperimen dengan MLflow, versi model dengan DVC, dan FastAPI.
100+ runs tracked02.EducationPendidikan
school
Master of Computer Science (S2)Magister Ilmu Komputer (S2)
Informatics EngineeringTeknik Informatika
Universitas Dian Nuswantoro GPA: 3.96/4.00
PeriodPeriode2024 — 2026 (ExpectedPerkiraan)
Research Focus: Integration of Artificial Intelligence (AI) in Intelligent Tutoring System based on adaptive pedagogical policies.
Core Courses: Advanced Machine Learning, Natural Language Processing (NLP), Multivariate Data Analysis.Fokus Riset: Integrasi Kecerdasan Buatan (AI) dalam Intelligent Tutoring System berdasarkan kebijakan pedagogis adaptif.
Mata Kuliah Utama: Machine Learning Lanjut, Pemrosesan Bahasa Alami (NLP), Analisis Data Multivariat.
school
Bachelor of Education (S1)Sarjana Pendidikan (S1)
Physics EducationPendidikan Fisika
Universitas Negeri Semarang GPA: 3.15/4.00
PeriodPeriode2018 — 2023
Focus Areas: Computational Physics, numerical methods, basic scientific programming, applied statistics, and pedagogical methodology.Bidang Fokus: Fisika Komputasi, metode numerik, pemrograman ilmiah dasar, statistika terapan, dan metodologi pedagogi.
03.SkillsKeahlian
Skills Proficiency DistributionDistribusi Tingkat Keahlian
Programming LanguagesBahasa Pemrograman
Core languages used to construct backend servers, machine learning pipelines, computational physics models, and mobile applications.Bahasa utama yang digunakan untuk membangun server backend, pipeline machine learning, model fisika komputasi, dan aplikasi seluler.
PythonSQLC++TypeScriptJavaScriptDartRustGoHTML / CSS
Frameworks & Libraries (AI-Focused)Framework & Pustaka (Fokus AI)
Toolkits and systems for generative AI (RAG, agentic LLM), deep learning, reactive SPA frontends, and robust backend APIs.Perkakas dan sistem untuk generative AI (RAG, LLM agen), deep learning, frontend SPA reaktif, dan API backend yang tangguh.
FastAPILangChainLlamaIndexHugging FacePyTorchvLLMMLflowDVCNext.jsReactNode.jsFlutterDockerKubernetes
Databases & SearchBasis Data & Pencarian
Relational databases, high-performance local engines, and vector search indexing (pgvector, Elasticsearch) for RAG.Basis data relasional, engine lokal berkinerja tinggi, dan indeks pencarian vektor (pgvector, Elasticsearch) untuk RAG.
pgvectorElasticsearchPostgreSQLSQLitebetter-sqlite3PineconeMySQLPrisma ORM
04.ExperiencePengalaman
codeAI Developer ExperiencePengalaman Developer AI
co_present
RolePeran
Full-Stack DeveloperFull-Stack Developer
Focus / PeriodFokus / PeriodeLectura — Academic Presentation EngineLectura — Engine Presentasi Akademikopen_in_new2025 — Present2025 — Sekarang
Built Lectura, a 3-mode academic presentation and whiteboard SPA using native ES Modules and a zero-dependency Node.js server.Membangun Lectura, sebuah SPA presentasi akademik 3-mode dan whiteboard menggunakan native ES Modules dan Node.js server tanpa dependensi.
<200ms slide compilations~85% faster lesson prepZero-dependency Node.jsModular architecture
psychology
RolePeran
ML & ITS EngineerML & ITS Engineer
Focus / PeriodFokus / PeriodeMaestro — Intelligent Tutoring SystemMaestro — Platform Intelligent Tutoring Systemopen_in_new2024 — Present2024 — Sekarang
Architected Maestro, an Intelligent Tutoring System using Flutter, Q-POMDP Reinforcement Learning, and local Markdown-based RAG.Merancang Maestro, sebuah Intelligent Tutoring System menggunakan Flutter, Q-POMDP Reinforcement Learning, dan RAG berbasis Markdown lokal.
HAKI CopyrightIEEE ISemantic 2026 submissionHallucination mitigation via response engineeringModular architectureFlutter + Firebase
analytics
RolePeran
Data AnalystAnalis Data
Focus / PeriodFokus / PeriodeData Analysis & Quantitative ResearchAnalisis Data & Riset Kuantitatifopen_in_new2023 — Present2023 — Sekarang
Developed automated ETL pipelines and R-based Network Meta-Analysis (NMA) engines using Pandas, NumPy, SQL, and MetaInsight. Performed statistical heterogeneity testing, funnel plot asymmetry analysis, and sensitivity analysis on 3 clinical trial datasets.Mengembangkan pipeline ETL otomatis dan engine Network Meta-Analysis (NMA) berbasis R menggunakan Pandas, NumPy, SQL, dan MetaInsight. Melakukan uji heterogenitas statistik, analisis asimetri funnel plot, dan analisis sensitivitas pada 3 dataset uji klinis.
3 clinical trial datasetsHeterogeneity testingSensitivity analysisReproducible NMA pipelineR + MetaInsight
monitoring
RolePeran
MTI Research AssistantAsisten Riset MTI
Focus / PeriodFokus / PeriodeTime-Series Forecasting with Multi-Head AttentionPeramalan Deret Waktu dengan Multi-Head Attentionopen_in_new2024 — 20252024 — 2025
Researched time-series forecasting using Multi-Head Attention architectures in PyTorch, tracked with MLflow, and published at ICAMIMIA 2025. Multi-Head Attention outperformed LSTM baseline with an RMSE delta of 88.34 on the test set — demonstrating statistical superiority in financial sequence modeling.Meneliti peramalan deret waktu menggunakan arsitektur Multi-Head Attention di PyTorch, dilacak dengan MLflow, dan dipublikasikan di ICAMIMIA 2025. Multi-Head Attention mengungguli baseline LSTM dengan delta RMSE 88.34 pada test set — menunjukkan superioritas statistik dalam pemodelan sekuens finansial.
Scopus publicationICAMIMIA 2025RMSE delta: 88.34PyTorch + MLflowModular architecture
gavel
RolePeran
Machine Learning & XAI ContributorKontributor Machine Learning & XAI
Focus / PeriodFokus / PeriodeJustExplain — Explainable AI Sentencing PredictorJustExplain — Prediksi Hukuman Pidana Terintegrasi XAIopen_in_new2025 — Present2025 — Sekarang
Collaborated on JustExplain, a court sentencing predictor utilizing ensemble models (XGBoost + Random Forest, 81% F1-score), SHAP explainability, and an interactive Streamlit dashboard. Trained on 22k+ documents from the Indo-Law corpus. Lead author: Jarot.Berkolaborasi membangun JustExplain, prediktor vonis hukuman pidana menggunakan model ensemble (XGBoost + Random Forest, skor F1 81%), eksplanasi SHAP, dan dashboard Streamlit interaktif. Dilatih pada 22k+ dokumen dari korpus Indo-Law. Penulis utama: Jarot.
81% F1 score22k+ docs — indo-lawEnsemble + SHAPModular architecture
psychology
RolePeran
NLP & MLOps EngineerNLP & MLOps Engineer
Focus / PeriodFokus / PeriodeLLM Fine-Tuning & MLOps PipelineLLM Fine-Tuning & MLOps Pipelineopen_in_new20252025
Fine-tuned LLM backbones (Hugging Face, GLM, LLaMA) using LoRA. Developed automated prompt testing and evaluation workflows with Ragas/TruLens, tracked via MLflow, and versioned via DVC.Melakukan fine-tuning backbone LLM (Hugging Face, GLM, LLaMA) menggunakan LoRA. Mengembangkan alur pengujian dan evaluasi prompt otomatis dengan Ragas/TruLens, dicatat dengan MLflow, dan diversi dengan DVC.
LoRA Parameter-efficient TuningRagas & TruLens EvaluationGit-based model tracking (DVC)MLflow Experiment RegistryOpenAI, Anthropic & GLM
alt_route
RolePeran
Backend & AI EngineerBackend & AI Engineer
Focus / PeriodFokus / PeriodeScalable FastAPI AI Service & GatewayScalable FastAPI AI Service & Gatewayopen_in_new20262026
Engineered a scalable AI inference service using FastAPI (Python) and orchestrated traffic routing on a Kubernetes cluster. Integrated hybrid pgvector and Elasticsearch RAG retrieval and deployed NeMo Guardrails for input/output security proxying.Membangun layanan inferensi AI berkinerja tinggi menggunakan FastAPI (Python) dan mengatur perutean trafik pada cluster Kubernetes. Mengintegrasikan retrieval RAG hibrida pgvector dan Elasticsearch serta menerapkan NeMo Guardrails untuk proxy keamanan input/output.
FastAPI & UvicornKubernetes & Dockerpgvector & Elasticsearch RAGNeMo Guardrails securityMicroservice routing
05.Featured ProjectsProyek Unggulan
Web
Lectura — Academic PresentationLive
An All-in-One Teaching Stage SPA built with native ES Modules, incorporating a Wiki-Study curriculum mode, a Reveal.js presentation engine with LaTeX/Mermaid integration, and an infinite Canvas whiteboard synced to SQLite.
▸ 3 Connected Modes▸ SQLite Sync▸ 5-Layer JS & 11-Layer CSS
StackNode.jsSQLiteReveal.jsMathJaxMermaid.jsesbuildPostCSS
Web
Maestro — Intelligent Tutoring SystemLive
An Intelligent Tutoring System (ITS) built on Flutter & Cloud Firestore, utilizing a Q-POMDP Reinforcement Learning engine and local Markdown RAG to deliver personalized pedagogical interventions via Groq/Gemini APIs.
▸ Flutter 3.41.7▸ Riverpod 2.6.1▸ HAKI Copyright
StackFlutterRiverpodQ-LearningFirebaseCloud FirestoreGroq/Gemini API
Academic
Udinus Thesis Template (LaTeX)Repo
A production-ready LaTeX thesis template for Udinus with one-command XeLaTeX compilation and arXiv citation integration — adopted by fellow students who were tired of fighting Word formatting.
▸ 2 GitHub forks▸ 1-command compilation▸ MCP arXiv integration
StackLaTeXXeLaTeXBiberArXiv APIMCP
Academic
Meta Analysis Data MedicRepo
ETL pipeline and Network Meta-Analysis (NMA) using R and MetaInsight to analyze antidepressant clinical trials and datasets, backed by Pandas, NumPy, and SQL.
▸ 3 clinical trial datasets▸ ETL pipeline▸ MetaInsight NMA
StackRPandasNumPySQLMetaInsight
Project
Machine Learning Modeling & Implementation
Multi-Head Attention architectures for stock market prediction published at ICAMIMIA 2025 (Scopus) — plus LSTM models deployed in production data pipelines for automated time-series forecasting.
▸ ICAMIMIA 2025▸ 100+ ablation runs▸ GPU cluster training
StackPythonTensorFlowPyTorchLSTMMulti-Head Attention
Project
JustExplain — Explainable AI Sentencing Predictor
A collaborative XAI research project predicting Indonesian court sentencing lengths using Random Forest + XGBoost with SHAP attribution and a Streamlit dashboard. Lead author: Jarot · Contributor: Praditya (ML pipeline, research design, data preprocessing).
▸ R² 0.81▸ MAE 7.03 mos▸ RMSE 11.87 mos▸ 22.6k+ cases
StackPythonScikit-LearnXGBoostSHAPMLflowStreamlitGitPythonlxmlPandasNumPySciPyStatsmodelsSeaborn
Project
LLM Fine-Tuning & MLOps Pipeline
A parameter-efficient LLM fine-tuning workflow (LoRA) integrated with comprehensive MLOps pipelines, prompt testing, evaluation frameworks (Ragas, TruLens), and model version tracking.
▸ 95% fewer parameters▸ Ragas & TruLens▸ MLflow
StackPythonHugging FaceLoRAMLflowDVCRagasTruLensOpenAI APIAnthropic APIGLM API
Web
Scalable FastAPI AI Service & Gateway
A high-performance AI inference service and security gateway built with FastAPI (Python), utilizing pgvector & Elasticsearch hybrid RAG, containerized with Docker, orchestrated on Kubernetes, and secured with LLM Guardrails.
▸ FastAPI▸ pgvector + Elasticsearch▸ Kubernetes
StackPythonFastAPIpgvectorElasticsearchDockerKubernetesNeMo GuardrailsLangChainLlamaIndex
Web
API Gateway — Enterprise PatternRepo
A zero-framework API Gateway implementing JWT authentication, sliding window rate limiting, and circuit breaker for microservice traffic management — built with native Node.js HTTP module.
▸ Zero-framework▸ JWT▸ Circuit Breaker
StackNode.jsSQLiteJWTDockerFly.io
06.Advanced Analytics & AI ResearchRiset AI & Analisis LanjutSelected PublicationsPublikasi Terpilih
Google ScholarORCIDNVIDIA DLI CertifiedSertifikasi NVIDIA DLIcopyrightIPR Reg. (Maestro)HAKI (Maestro)
Research Focus & Interest DistributionFokus Riset & Sebaran Bidang Minat
description
Accepted — Scopus-Indexed (2026)Diterima — Terindeks Scopus (2026)Stock Market Prediction and Time-Series Forecasting using Multi-Head Attention
P. Wicaksono, R. A. Pramunendar — ICAMIMIA 2025. (Scopus)
Deep LearningTime-SeriesAttention MechanismsScopus
smart_toy
Under Review (IEEE Conference, 2026)Sedang Ditinjau (Konferensi IEEE, 2026)MAESTRO: A Hybrid Q-Learning and Large Language Model Architecture for Intelligent Tutoring Systems with Adaptive Pedagogical Policies
P. Wicaksono, P. N. Andono, Pujiono — 2026.
07.Latest ThoughtsCatatan Terbaru
rss_feed
rss_feed
08.Contributions GraphGrafik Kontribusi
09.08.ContactKontak
Working on an AI or engineering project? Open for technical discussions and research collaboration.Sedang mengerjakan proyek AI atau engineering? Terbuka untuk diskusi teknis dan kolaborasi riset.Open for research collaboration — thesis discussions, joint publications, or exploring AI in Education together.Terbuka untuk kolaborasi riset — diskusi tesis, publikasi bersama, atau mengeksplorasi AI dalam Pendidikan.