Issue №01 · July 2026 · Bandung, Indonesia
Building AI that works in the real world.
An AI engineer who ships systems people actually use — from the LLM framework my team builds on every day, to the knowledge-graph stack a government client depends on. Currently with PT. Indonesia Indicator.
Less magic. More boring infrastructure that holds.
I build AI that earns its keep — observable, debuggable, and cheap enough to leave running. The interesting work is rarely the model; it's the plumbing around it, the failure modes, and the question of what should happen at 3 a.m. when something upstream changes.
Right now that means LLM frameworks the whole team builds on, cost-tuned inference pipelines, and knowledge-graph stacks where hallucination is a regulatory problem, not a UX one. I'm happiest when the work disappears into the product — when a teammate stops thinking about “the AI” and just uses it.
Outside that, I'll cheerfully ship a clean React app, a mobile build, or a half-baked pixel-art weekend project. The breadth keeps the depth honest.
Working knowledge
AI & LLM
- LLM Agents
- A2A
- LangChain
- OpenAI SDK
- MCP
- RAG
- Knowledge Graph
- Fine-tuning
- OpenTelemetry
- Langfuse
Backend & Cloud
- Python
- FastAPI
- Flask
- Docker
- Kubernetes
- AWS
- GCP
- PostgreSQL
- Memgraph
- CI/CD
ML & Data
- TensorFlow
- PyTorch
- Computer Vision
- CNN
- Sentiment Analysis
- Social Network Analysis
Shipping rhythm
A rough shape of the last year — hover or tab through a day for a read on how busy it was. Most of the actual commits live in private company repositories and don't show up on a public profile.
Where the lessons came from.
Jan 2025 – Present
Current
Tangerang Selatan, Indonesia
AI Engineer/Researcher · PT. Indonesia Indicator
- Built an LLM framework that our whole team now uses — supports multiple AI providers and just works out of the box (saved us months of dev time)
- Made our AI infrastructure way cheaper (75% cost reduction) while keeping things reliable with proper observability
- Trained a custom image model for architectural designs — what used to take hours now takes minutes
- Created AI agents that write reports and presentations, so teams can focus on the actual work
- Built a knowledge graph system for government clients where accuracy really matters — hallucination went down to basically zero
- Working on a no-code app builder where you just describe what you want and it builds it
- LLM Agents
- RAG
- Kubernetes
- Langfuse
- Graph RAG
- MCP
Jul 2021 – Oct 2024
Bandung, Indonesia
Data Scientist · UNIKOM CODELABS
- Built SociaLabs — a social media analytics tool that actually makes sense of Twitter data with AI-powered insights
- Made Agrimate, an app that helps farmers spot crop diseases with 95% accuracy (pretty proud of this one)
- Created MainChick, a smart poultry management system — turns out chickens need data too!
- Got really good at shipping ML models to production without breaking things
- Python
- TensorFlow
- Docker
- AWS
- GCP
- FastAPI
Feb 2023 – Jul 2023
Indonesia
Machine Learning Engineer · Bangkit Academy
- Got into this pretty competitive program (20k+ applicants) — felt lucky to learn from Google, Tokopedia, and Gojek folks
- Built a pet adoption app that matches you with the right pet based on your lifestyle (the breed recognition hits 98%!)
- Learned to work across teams and ship ML models that don't break in production
- Got my TensorFlow certification — turns out I actually know what I'm doing 😅
- TensorFlow
- CNN
- Docker
- GCP
- Android
Things shipped, in production, used by real people.
A short list, ordered by what taught me the most. Read any title for the full case.
- №01
Agrimate
An app that helps farmers take better care of their crops — snap a photo and it flags disease, plus smart watering and price predictions.
- CNN
- TensorFlow
- IoT
- №02
MainChick
A smart assistant for poultry farmers — monitors barn conditions and early signs of disease, with a chatbot that genuinely knows its chickens.
- Machine Learning
- IoT
- Chatbot
- №03
Peaky Blinder
Eye check-ups through your phone camera — the AI spots early signs of diabetic eye disease, with a chatbot to answer health questions.
- Azure Vision
- LLM
- Healthcare
- №04
OPet
Finding the right pet match — OPet learns what fits your lifestyle, shows adoptables nearby, and identifies breeds from a photo.
- CNN
- Image Recognition
- Maps API
- №05
SociaLabs
Making sense of Twitter chaos — what people talk about, how they feel, and who drives the conversation.
- NLP
- Social Network Analysis
- Sentiment Analysis
More on GitHub →
Recognition along the way.
A handful of competitions and certifications that taught me to ship under pressure.
- №012023 & 2024
Global Top 100 Finalist
Google Solution Challenge
- №022022
Top 10 of 625 Teams
Microsoft Imagine Cup
- №032023
1st Runner-Up + Audience Choice
COMPFEST AI Innovation Challenge
- №042023
National Finalist
Gemastik XVI Software Engineering
- №053rd sem – graduation
Rector's Scholarship
UNIKOM
- №062023
TensorFlow Developer Certified
Google
Writing — mostly on AI in production.
Field notes from shipping models, agents, and the messy infrastructure underneath — written here, with some cross-posted from Medium.
The full archive lives on Medium →
Write to me — I read everything.
I'm up for full-time roles, contract work, advisory, or a good long conversation about where AI is going wrong. Email is the fastest path.
- Phone
- +62 85295451122
- Based
- Indonesia
- Elsewhere
- GitHub →LinkedIn →Email →