02 / What I do

Work & projects.

A decade of building intelligent autonomous systems — translating machine-learning research into operational software, from hardware-level TensorRT inference to enterprise agentic business systems.

21
Public repositories
5
Engineering domains
10+
Years shipping
Experience

Where I've worked

Senior Computer Vision Engineer & CV Team Lead
Sigmawave AI Pte. Ltd. — Singapore
Sep 2026 – Present · Remote / Hybrid
  • Rebuilt and optimized a multi-camera person re-identification and cross-camera tracking system for edge deployment; lead the CV engineering team across model design, optimization, and production MLOps.
  • Optimize deep-learning models for inference efficiency, latency, throughput, and hardware constraints on edge targets.
  • Own task allocation, technical standards, and code reviews for the CV team; coordinate cross-functional workflows to hit product milestones.
  • Establish experiment tracking, model versioning, and CI/CD for production reliability; define confidence thresholds and fallback behaviours for safe real-world deployment.
AI Architect & Technical Lead
V-TEKI
Jul 2026 – Present · Jakarta · Remote
  • Set technical direction for enterprise AI delivery across LLM, agentic, and computer-vision workstreams — defining target architectures and reviewing designs and source code to production standard.
  • Drive MLOps and deployment standards (experiment tracking, model versioning, CI/CD) and lead architecture decisions from proposal through client delivery.
  • Establish best practices for AI engineering, deployment, and MLOps; contribute reusable frameworks, standards, and documentation.
  • Mentor AI engineers and cross-functional teams; support delivery through proposal reviews, solution architecture, and client discussions.
Technical Mentor — Garuda Hacks 7.0
Garuda Hacks
Jul 2026 · Remote
  • Provided technical guidance on system design, AI/ML integration, and agentic-workflow deployment to hackathon teams under rapid-prototyping constraints.
  • Advised on technical viability and early-stage architecture to help teams avoid structural technical debt.
Facilitator — Google Skills Arcade 2026
Dicoding Indonesia
Jul 2026 – Present · Remote
  • Mentor national tech talent through Google Cloud hands-on labs on cloud architecture and infrastructure.
  • Facilitate troubleshooting discussions and community networking within the Google Skills developer ecosystem.
Google Prompt Engineering Trainer (VILT)
The Smartbridge International — Last Mile Indonesia
Feb 2026 – Jul 2026 · Remote
  • Lead trainer for a 30-hour VILT on advanced Prompt Engineering for developer cohorts in Indonesia's national upskilling program.
  • Taught CoT, ToT, and Self-Consistency reasoning; the RTCFC prompt-architecture framework; and model parameters (Temperature, Top-P, Top-K).
  • Covered agentic workflows on the Gemini ecosystem, cost-accuracy trade-offs, tokenization, and red-teaming for enterprise AI safety.
AI Image Data Contributor (Freelance)
SoftAge Information Technology Limited
Feb 2026 – Jul 2026 · Remote
  • Contributed to the SrotPix Image Data Collection Project — generated high-quality structured visual datasets for AI model training across Character Consistency, Nested Objects, and Object Interactions.
  • Ensured strict adherence to dataset QA guidelines: metadata consistency, lighting continuity, and precise spatial framing.
  • Gained hands-on experience in the foundational data-gathering phase of the Generative AI model development lifecycle.
Freelance Software & AI Engineer
PostWork AI
Oct 2025 – Jul 2026 · Remote
  • Developed, fine-tuned, and evaluated models prior to production deployment.
  • Built bespoke automation and ML-integration solutions across diverse remote engagements.
Lecturer & Researcher — Electrical Engineering
Soegijapranata Catholic University
May 2014 – Present · Semarang
  • Designed and delivered courses in Embedded Systems, AI/ML Applications, and Autonomous Robotics to 20–40 students per semester for 10+ years.
  • Supervised 10+ concurrent UAV / AUV / AGV / humanoid research teams annually, from hardware design to deployment.
  • Architected real-time CV pipelines for autonomous UAVs and quadrupeds (OpenCV, PyTorch, Raspberry Pi).
  • Research: Google Scholar 40 citations, h-index 4 · Scopus 18 citations, h-index 3.
General Manager & Principal Adviser — Robotics Team
Soegijapranata Catholic University
Aug 2014 – Present · Semarang
  • Founded and led the robotics team across 10+ national competitions spanning UAV, AUV, AGV, and humanoid platforms.
  • 1st place nationally — AEROCREATION 2017 (ITB); 5th nationally — Indonesian Aerial Robotics Contest 2016.
  • Multiple top-10 national finishes (2015–2024) across VTOL, Underwater, SAR, and Humanoid categories.
Portfolio

Featured projects

Sectors Hackathon 2026 · Live demo

Neraca Fakta — IDX Stock-Claim Fact-Checker

Stock tips spread fast in Indonesia — forwarded chats, Telegram groups, screenshots. Neraca Fakta pulls every numeric claim out of a tip and checks it against official IDX financial data via the Sectors API. The LLM (Claude, OpenAI or Gemini) only extracts typed claims; a deterministic engine assigns each verdict against fixed tolerances, so the model never decides the outcome. Web app plus Telegram and WhatsApp bots, bilingual, with a 98-test suite.

PythonFastAPIReact / TypeScriptClaude / OpenAI / GeminiSectors APITelegram / WhatsAppVercel
Try the live demo ↗    View the repository ↗
OpenClaw Agenthon 2026 · Best Payment Use Case track

LapakPintar — Autonomous Business OS

Five QwenPaw-orchestrated agents for Indonesia's 65M SMEs (Team NexaPaw): competitor price monitoring (Playwright), demand forecasting (Qwen3 + RAG), content generation, DOKU payment reconciliation, and daily reporting — running 24/7 with zero human intervention.

QwenPaw / AgentScopeQwen3-14BPlaywrightDOKU REST APIFastAPIRedisDocker
View the repository ↗

⚡  Edge AI & High-Performance Computing

5 repos
Visitor Counting Console
2026 · Active
Footfall analytics from existing CCTV — batched YOLO11 detection, per-camera ByteTrack, trigger-line entry/exit counting, and SigLIP 2 zero-shot gender and adult/child demographics voted over each track. FastAPI console with live overlays for up to six cameras.
YOLO11ByteTrackSigLIP 2FastAPIOpenCVCUDA / MPS
crosscam-reid
2026
Cross-camera person re-identification from scratch — per-view IoU tracking, colour-histogram or ResNet50 appearance embeddings, and EMA-prototype gallery fusion so one person keeps one ID across cameras. The self-scoring synthetic demo runs fully offline.
PythonOpenCVNumPytorchvisionRe-ID
PerceiveReason · Apple Silicon
2026 · WWDC 2026
On-device Perceive → Reason pipeline — Core ML + Vision on the Neural Engine for perception, a swappable LanguageModel (Apple Foundation Models or Claude) for reasoning. MobileNetV3 conversion with int8 quantisation plus a SwiftUI reference app.
Core MLcoremltoolsSwiftSwiftUIFoundation ModelsClaude API
PPE_Detection_TensorRT
2024 – Present · Active
Enterprise PPE monitoring on NVIDIA Jetson — TensorRT FP16 with zero mAP loss, Forward Hooks for Layer-Collapse mitigation, dynamic batching, and an async FastAPI gateway handling 15×1080p streams at ≤66.7 ms.
PyTorchTensorRTYOLOv11FastAPIJetsonC++
kyc-vision-system
2025 – Present · Active
Production KYC verification — high-concurrency Go 1.22 gateway (JWT, rate limiting) + Python CV service: Indonesian KTP OCR, YOLOv8 face detection, image-quality gating, and selfie liveness. Full CI/CD.
Go 1.22PythonFastAPIOpenCVYOLOv8Tesseract

🧠  Generative AI & Agentic Systems

8 repos
RouteReason
2026
Plain-language delivery routing — GPT-5.6 parses a described run into a schema-validated routing problem, a pluggable solver (Clarke-Wright reference, OR-Tools ready) plans it, and GPT-5.6 explains the tradeoffs alongside an interactive map.
GPT-5.6OR-ToolsPythonSchema contracts
AD Applicability Extraction
2026 · Aviation · Complete
Reads FAA / EASA Airworthiness Directive PDFs and answers “is aircraft X affected by AD Y?” — deterministic clause isolation, schema-bound LLM extraction, multi-run stability checks, and grounding validation against the source text.
PythonPydanticpdfplumberGemini APILLM extraction
EduSense AI
2026 · Full-stack
Adaptive learning platform — a Claude-powered tutor and quiz generator, course library, learning analytics, and XP / badge gamification with JWT auth.
FastAPIClaude APIReact 18ViteSQLAlchemyTailwind
Enterprise Agentic Pipeline
2026 · Human-in-the-loop
Next.js + GPT-4o pipeline for multi-step document workflows — Zod-enforced structured outputs, tenant-isolated RAG context, and a self-scored confidence gate that hands low-confidence cases (below 0.92) to human review.
Next.jsTypeScriptGPT-4oZodRAG
Enterprise AI Context Engine
2024 – Present
Unifies AML/fraud detection, credit-risk underwriting, and marketing analytics via LangGraph's cyclic multi-agent orchestration, MCP tool registration, and a human-readable explainability layer for compliance.
LangChainLangGraphMCPGemini APIScikit-LearnGCP
synapse
2024 · Full-stack
AI meeting assistant on the Claude API — transcript → executive summary, action items, decision log, key topics, and a follow-up email draft. FastAPI + static frontend via Nginx, Docker Compose.
FastAPIClaude APINginxDockerGitHub Actions
gemini-chatbot-v2
2024
Full-stack Node.js chatbot on the Gemini API with token-by-token streaming and multi-turn context management.
Node.jsExpressGemini APIJavaScript
langgraph_agent_test
2025 · Experimental
Cyclic, stateful multi-agent experiments on Google Cloud Platform — exploring LangGraph graph architectures for autonomous planning and tool use.
LangGraphGCPPython

🏦  Document AI, Identity & Enterprise Architecture

3 repos

🤖  Robotics & Autonomous Systems

3 repos

🔬  Scientific CV, Web & Tools

5 repos

🛰  Latest on GitHub

live

Loading the latest repositories…