Athilla Zaidan

itb cs, class of 2028. fullstack by training, ml by obsession.

i build the boring infrastructure around models (evals, pipelines, judges) because that’s where the interesting failures hide. right now i’m benchmarking llms as an ai engineer intern, which mostly means finding creative ways for models to be wrong, then writing it down with error bars.

i’m drawn to anything with a lid on it. i’ve pointed cameras at public transport and asked a model what went wrong, taught a classifier to tell fish apart (then taught a judge to tell me when the explanations were lying), read thousands of citizen complaints in indonesian so no one has to, and graded vegetables by how tired they look. i also write the unglamorous stuff: registration systems that survive real payments, a recommender, a strategy game with plugins.

a few hackathon podiums so far, mostly from starting the next build before the last one cooled down. currently learning mlops, so the pipelines can finally run without me watching them.

Work experience

  1. AI Engineer Intern, GDP Labs

    Jun 2026 – now

    • Built a full-stack LLM benchmarking platform (FastAPI, Alembic, PostgreSQL, React) that moved results from spreadsheets into a 532-deployment model registry with run tracking and a human-review dashboard.
    • Benchmarked 35+ frontier LLMs from 15 providers, including GPT-6, Claude 5.5, and Gemini 3.8, on RAG, AI-agent, vision, and LLM-as-a-judge tasks in both reasoning and non-reasoning modes.
    • Engineered a multimodal LLM judge that grades the charts an AI agent produces, and automated per-model accuracy and cost reports after every run, replacing a manually assembled comparison table.
    • Modernized the evaluation codebase by migrating 8 benchmark pipelines to a shared interface and Poetry to uv, removing over 4K lines of legacy code while preserving leaderboard output.
  2. Head of IT, Fullstack, StudentxCEOs Grandsummit 15th

    Feb 2026 – now

    • Led a team of seven to ship registration for 280+ teams, 1,500+ file uploads and Rp20M+ in revenue, with Supabase and Google Drive failover storage.
  3. Technology Development Intern, Inkubator IT

    Oct 2025 – Mar 2026

    • Delivered a client platform in Next.js from Figma, with a weighted Brand DNA quiz, radar charts and a serverless news digest.
  4. Frontend Developer, TEDxITB 9.0

    Oct 2025 – May 2026

    • Built the event site and its ticket commerce flows with server components, OpenGraph metadata and a generated sitemap.
  5. Fullstack Developer, IT Aku Masuk ITB 2026

    Oct 2025 – Feb 2026

    • Built FindMi, a merchandise store that took IDR 200M+, and an event platform for 10,000+ users with RSVP and QR tickets.

Projects that made it out alive