~/about⎇ main✓❯ls -la
clariant background.md clariant interests.md clariant working-on.md
The short version,
then the longer one.
Background.
I'm Clariant Benedictus Tan, a Computer Science student specializing in Intelligent Systems. I got pulled into this field by the moment large language models stopped being clever toys and started being substrate — when "you can call them in a loop with tools" turned out to actually mean something.
I learn by building. My academic projects skew toward applied ML — CNNs for traffic-sign recognition under realistic deployment constraints, regression pipelines for housing prices, end-to-end web apps that use AI rather than just demo it.
What I care about.
Where my attention actually goes: multi-agent orchestration, RAG systems that hold up on real corpora, and getting local LLMs to do useful work without phoning home to a frontier API.
The reason these three things sit together: each one is about making AI compose — with tools, with itself, with constrained hardware. Frontier models that solve one prompt at a time aren't the interesting part anymore. The interesting part is the system you build around them.
Right now.
Interning at Dewaweb, owning a chunk of their AI surface — building a RAG-grounded copilot for the support team, running the backend for ScaleCon Surabaya 2026, evaluating agentic CLIs and local models for production adoption.
On the side: poking at Hermes and OpenClaw as agent harnesses, benchmarking Qwen 3.5/3.6 and Gemma 4 on workloads that actually matter to me. Live logs at ❯ ~/now .