Skip to content
View winstonkoh87's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report winstonkoh87

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
winstonkoh87/README.md

Hi, I'm Winston Koh 👋

Last updated: 5 April 2026

AI Systems Engineer | Creator of Project Athena | Singapore 🇸🇬

I build agentic AI infrastructure that turns generic LLMs into personalised operating systems — own the state, rent the intelligence.

Website: winstonkoh87.com


🎯 What I Build

Project What It Does Engineering Highlights
Athena Open-source cognitive augmentation layer — persistent memory, structured reasoning, full data ownership across any AI model (Gemini, Claude, GPT) • Hybrid RAG pipeline (BM25 + semantic + knowledge graph + RRF fusion)
• 85% recall @ $0 infrastructure cost (Supabase free tier)
• 150+ reusable protocols, 66 slash commands, 30+ skills
• Conditional skill activation (~40-60% token savings)
• 7 IDE integrations, MIT licensed
Portfolio 11-page performance-first portfolio with 26 articles and 6 live demos • Astro 5.0 + Tailwind CSS + React
• Zero-JS-first architecture, 99/100 Lighthouse

🧠 Tech Stack

Languages & Frameworks

Python TypeScript Astro JavaScript HTML5 CSS3

AI & Data

Gemini Claude OpenAI Supabase PostgreSQL

Tools & Infrastructure

Git GitHub Actions VS Code Cloudflare


📊 Athena v9.6.6 — By the Numbers

A platform-agnostic cognitive augmentation layer. Own the state. Rent the intelligence.

Metric Value What It Means
Sessions 1,500+ Continuous context across 90+ days of bilateral use
Protocols 150+ unique Open-sourced decision frameworks (reasoning, risk, execution, research)
Hybrid RAG 85% recall BM25 + semantic + knowledge graph + FlashRank reranking
Skills 30+ clustered Cognitive Cluster architecture — co-activated skill pipelines with conditional activation
Case Studies 430+ Documented friction → solution patterns with empirical outcomes
Scripts 650+ Full automation stack (boot, shutdown, search, sync, hooks)

Key Engineering:

  • Hybrid Search: pgvector + BM25 keyword + knowledge graph + RRF fusion — outperforms basic RAG by 48pp (92% vs 44%)
  • Conditional Skill Activation: Path/topic-triggered dormant skills reduce prompt bloat by ~40-60% (Protocol 530)
  • Multi-Agent Orchestration: Coordinator synthesis discipline with anti-delegation enforcement and token budgeting
  • Atomic Writes: POSIX-compliant data safety for all memory operations
  • Privacy Pipeline: block_secrets.py git hook + PII regex scrubber + public/private repo guard
  • Semantic Cache: LRU with disk persistence, cosine matching, and Supabase delta sync
  • Zero Infrastructure Cost: Runs on Supabase free tier + local compute. No cloud bills.

⚡ Capability Stack

Capability Evidence
RAG Pipeline Engineering Production hybrid search: BM25 + semantic + graph + RRF fusion. 85% recall, $0/month infra
Agentic AI Systems 150+ protocols, 66 workflows, 30+ skills — full agent lifecycle (boot → work → shutdown)
Multi-Agent Coordination Parallel worktree orchestration, coordinator synthesis, conditional skill activation
Full-Stack Web Development 5 production sites, Astro/React, zero-JS-first architecture
AI Consulting Active client engagements — diagnostics, AI integration strategy, workflow automation
Technical Writing 26 articles, 9.8K+ views — clear communication of complex systems

✍️ Recent Writing


💞️ Open to Collaborate

Looking for pragmatic builders who:

  • Ship fast, iterate, break things at 70% readiness
  • Value robustness over cleverness
  • Are comfortable with async communication

📬 Reach out: winstonkoh87@gmail.com or DM on LinkedIn


📫 Connect

Portfolio LinkedIn Email


"The best way to predict the future is to build it."

Pinned Loading

  1. Athena-Public Athena-Public Public template

    The Linux OS for AI Agents — Persistent memory, autonomy, and time-awareness for any LLM. Own the state. Rent the intelligence.

    Python 452 62