AI Agents & Automation
AI agents and workflow automation that run inside your own systems.

(01)What we build
Start with an AI agent, with process automation, or with a product built end to end. Same team either way.
AI agents and workflow automation that run inside your own systems.
Websites, web apps, SaaS products, and internal dashboards through to production.
Assessment, roadmap, training, and implementation advisory for AI work.
Data pipelines, system integration, API work, and backend platforms.
(02)Selected work
Our own products, open-source tools, and client work — some anonymized under NDA. The scope and the technical results are real and checkable.
(03)Engineering notes
Dated technology updates, tutorials you can follow, and pattern analyses from our work. The four most recent notes.

DeepSeek open-sourced its super-agent prototype on October 31, 2026: a proactive coding agent on Harness. How it works, how it compares, and the limits.
Read the note
On October 21, 2026, Node.js shipped 11 CVE fixes and PHP shipped two critical RCE fixes the same day. What changed, and the patch order to follow.
Read the note
Prompt changes silently break downstream agent tool calls. A guide to building deterministic eval harnesses, trajectory metrics, and CI regression testing.
Read the note
GraphRAG enables multi-hop reasoning across documents, but indexing cost and latency are high. Technical comparison with Hybrid RAG and trade-offs.
Read the note(04)Choose your starting point
The figures below are starting points, not a final quote. We lock the exact scope after the discovery session.
AI agents and workflow automation that run inside your own systems.
Websites, web apps, SaaS products, and internal dashboards through to production.
Assessment, roadmap, training, and implementation advisory for AI work.
Data pipelines, system integration, API work, and backend platforms.
All prices on this page are estimates for reference. The final figure can be lower or higher, adjusted to your requirements, complexity, and scope of work.
The rupiah figure is indicative, converted from USD at a rate we review periodically.
(05)Who we work with
Industries
(06)How we work
Every stage has an output you can look at and respond to, not a status report.

Week 1
We study the process, the data, and the constraints. The output is one scope document with the assumptions written down, not a verbal estimate.

Weeks 2 to 6
Development runs in weekly cycles. Every week there is a version you can try, not one demo at the end.

Weeks 6 to 8
We connect the system to the tools you already use, then deploy with monitoring switched on from day one.

Ongoing
After launch we keep watching performance, model cost, and the cases that fail, then fix them in the same cycle.
(07)Questions, answered
If your question is not here, send it over on WhatsApp or by email.
It depends on the engagement. An AI assessment runs four weeks. One AI agent workflow through to production is typically six to twelve weeks. A company site is four to eight weeks. A SaaS product to a first usable version is usually three to six months. Every proposal names dates per stage, and you can open the staging environment from the first sprint rather than only at the end.
Fixed price per stage, not hourly billing. Once the scope document is agreed we send one number for that stage, with payment terms split across kickoff, midpoint, and handover. If the requirement changes mid-project we write it up as a scope change with its own number before any work starts on it. Contracts are written in USD and can be paid in rupiah at the rate on the invoice date; Indonesian clients receive an invoice with the applicable tax documents.
Yes. We are based in Jakarta (WIB, UTC+7) and work with clients across Southeast Asia, Australia, Europe, and North America. Our hours overlap fully with Singapore, Australia, and East Asia, and overlap with Europe during our morning. For North American clients we hold a late-afternoon WIB block for live calls and run the rest asynchronously through written updates and recorded demos. Project working language can be English or Indonesian.
You do. All source code, infrastructure configuration, prompts, and evaluation sets are handed over in a repository and cloud account held in your company name, and ownership transfers on final payment. Your data stays in your own infrastructure. Foundation model weights remain the property of their provider (OpenAI, Anthropic, Google, or an open model under its own licence), but every fine-tune, index, and embedding produced from your data is yours. We do not use client data to train anything.
Every engagement ends with a handover: technical documentation, an operational runbook, and one recorded knowledge transfer session with your team. After that there is a 30 day warranty period covering bug fixes within the delivered scope at no extra cost. If you want us to stay involved, a monthly support retainer covers monitoring, dependency updates, and a block of development hours. The retainer is optional and is never a condition of handover.
Yes, and it is the most common pattern we run. We join your repository, your code standards, and your review process rather than imposing ours. If your team stands on Java, .NET, Python, or Go, we work there. In this kind of engagement we usually pair with your engineers so the knowledge stays after we leave. If the existing stack genuinely does not suit the new requirement, we explain the technical reason and the alternatives, and the decision stays with you.
Send a message through the contact form or WhatsApp with a short description of the problem. We reply within one business day and schedule a 45 minute call to understand your process, your systems, and your deadline. That call is free. After it we send one document covering the problem as we understood it, the approach we propose, the stages, and the price per stage, usually within three to five business days. If the approach fits, we start with the first stage. If we think your problem does not need us, we say that too.
Jakarta · Indonesia · Worldwide
You get a reply from the team that would do the work, not from a sales desk.