Indonesia, ID
Open for work
Systems Builder
Systems + Products
DWIKYCANDRA
I design and ship systems that make complex work clear: from operational tools and data products to interfaces that help people decide and move. Research → build → ship.
I build systems that make complex work clearer: products, interfaces, and infrastructure that help people understand, decide, and ship.
I'm Dwiky, a systems-minded builder working across product interfaces, data, automation, and infrastructure. I care about the seams: how a model becomes a workflow, how a dataset becomes a decision, and how a tool behaves when the conditions are not ideal.
My work moves between operational tools, knowledge systems, and interfaces for real people. I like software that stays legible under pressure: clear states, honest data, useful defaults, and an escape hatch when the happy path breaks.
I build and maintain products such as Luminary Memory, ZeroCode, KyDev, and Kasir Pintar. I also lead JWIS, a civic operations system that turns fleet, forecast, and field signals into a clearer decision loop.
Intensive bootcamp on LLM architecture, prompt engineering, fine-tuning, and AI integration in web applications. Capstone: deployed semantic search feature serving 500+ active students.
Visit program- 01LLM architecture & tokenization internals
- 02Production prompt engineering patterns
- 03RAG pipeline design with vector stores
- 04AI feature deployment for 500+ users
Systems, products, and interfaces built to make complex work clearer.
A command center for the Dinas Lingkungan Hidup (DLH) DKI Jakarta waste case, connecting fleet supervision, route decisions, forecasting, and field dispatch in one operational loop.
- Command-center workflow for DLH DKI Jakarta to supervise fleet movement and waste operations
- Fleet map with route deviation signals and a live operational view
- A* and OSRM route recommendations with queue-aware dispatch planning
- Waste-volume forecasting and resource planning across 42 Jakarta kecamatan
- Manager-to-field dispatch flow with confirmation and audit context
- Fleet, TPA queue, forecast, and field signals stay in one operational context
- Integrated planning translates forecast signals into fleet, crew, and dispatch decisions
- Prophet and XGBoost models expose forecast metrics, factors, and evaluation scope
- Real, modeled, simulated, and fallback data are labeled separately
Everything else, still worth finding.
3 results
Selected signals from work, study, and shipping.
“Delivered production-ready code with 95% test coverage. Consistently meets deadlines with clean, maintainable solutions.”
“Led frontend architecture decisions that reduced load time by 65%. Strong technical leadership and clear communication.”
“Exceeded project requirements. Delivered 2 weeks early with comprehensive documentation and 99.5% uptime.”
“Intuitive UI with flawless mobile responsiveness. Performance optimizations made the app feel instant.”
The tools change. The systems thinking stays.
Intelligent Systems
(01)- Memory Systems
- Retrieval & RRF
- FastEmbed / ONNX
- LangGraph
- Knowledge Graphs
- Data Pipelines
- Model Evaluation
- RAG Systems
Infrastructure
(02)- Linux
- Rust
- Tauri
- Python
- Bash
- Systemd
- SQLite
- Btrfs
Product Interfaces
(03)- TypeScript
- React 19
- Next.js 16
- Tailwind CSS
- Vite
- Motion design
Data & Backend
(04)- Python
- Node.js
- PostgreSQL / pgvector
- SQLite / FTS5
- FastAPI
- Prisma
A visible trail of the work
Keep shipping. Let the trail speak.
Public GitHub activity across the systems, products, and experiments in the archive.
Open GitHub