Name
https://github.com/juliusbrussee/caveman
GitHub Handle
@JuliusBrussee
Tell us about yourself
I'm 19 and building AI-powered tools that solve real workflow problems.
Currently working on two systems:
Stacklink — An enterprise RAG platform that acts as organizational memory, integrating with Slack, Notion, and GitHub. Built for the TU Eindhoven innovation contest with a full TypeScript/Next.js stack, pgvector semantic search, and hybrid retrieval architecture. 96k+ lines of production code with comprehensive testing.
Caveman — An open-source token compression project for LLMs and agentic systems. The premise is simple: strip articles, filler, and politeness from prompts and cut ~65% of your tokens for the same output. What started as a single Claude Code skill grew into a full ecosystem (CaveMem for semantic memory, CaveKit for spec-driven parallel development, and Cave Gemma, a Gemma 4 31B LoRA fine-tune that bakes the style into the model weights). 62,000+ GitHub stars, Top 300 most-starred repository on GitHub, and circulating internally at Nvidia.
Featured on BNR Nieuwsradio — the Netherlands' largest business and tech news network — as a piece of Dutch open-source infrastructure.
I serve on the Education Committee at Leiden University's LIACS, representing students in curriculum development for the BSc Data Science & AI program.
Technical focus: Full-stack development (Swift, TypeScript, Python), AI/ML systems, RAG architectures, native macOS/iOS, and building products that people actually want to use.
Project Name
Caveman
Project Repo Link
https://github.com/juliusbrussee/caveman
Stream Date
Dates
08-21-2026
Twitter URL
https://x.com/julius_brussee
LinkedIn URL
https://www.linkedin.com/in/julius-brussee/
Additional Information
No response
Name
https://github.com/juliusbrussee/caveman
GitHub Handle
@JuliusBrussee
Tell us about yourself
I'm 19 and building AI-powered tools that solve real workflow problems.
Currently working on two systems:
Stacklink — An enterprise RAG platform that acts as organizational memory, integrating with Slack, Notion, and GitHub. Built for the TU Eindhoven innovation contest with a full TypeScript/Next.js stack, pgvector semantic search, and hybrid retrieval architecture. 96k+ lines of production code with comprehensive testing.
Caveman — An open-source token compression project for LLMs and agentic systems. The premise is simple: strip articles, filler, and politeness from prompts and cut ~65% of your tokens for the same output. What started as a single Claude Code skill grew into a full ecosystem (CaveMem for semantic memory, CaveKit for spec-driven parallel development, and Cave Gemma, a Gemma 4 31B LoRA fine-tune that bakes the style into the model weights). 62,000+ GitHub stars, Top 300 most-starred repository on GitHub, and circulating internally at Nvidia.
Featured on BNR Nieuwsradio — the Netherlands' largest business and tech news network — as a piece of Dutch open-source infrastructure.
I serve on the Education Committee at Leiden University's LIACS, representing students in curriculum development for the BSc Data Science & AI program.
Technical focus: Full-stack development (Swift, TypeScript, Python), AI/ML systems, RAG architectures, native macOS/iOS, and building products that people actually want to use.
Project Name
Caveman
Project Repo Link
https://github.com/juliusbrussee/caveman
Stream Date
Dates
08-21-2026
Twitter URL
https://x.com/julius_brussee
LinkedIn URL
https://www.linkedin.com/in/julius-brussee/
Additional Information
No response