Media Kit
Resources for press, investors, and partners
Company Overview
One-Liner
Hire a 10x AI Automation Engineer to kill your busywork
Founded
Founded in November 2023 in Germany
Location
Mission
Liberate the world from busy work that no one wants to do with the highest quality and at the cheapest price.
Description
PUNKU.AI creates, improves, and shares AI Automations for you. No coding, just english! Transform your business with intelligent automation.
What Does PUNKU Mean?
PUNKU (pronounced "POON-koo") comes from the Quechua and Aymara word meaning "portal" or "door". The name is inspired by Puma Punku, an ancient archaeological site in Bolivia at 3,960 meters above sea level, where precisely carved stone gateways were believed to open passages into new dimensions.
Just as those ancient portals transformed who could walk through, PUNKU.AI serves as a modern gateway—opening access to AI tools that once required specialists or engineers, democratizing who gets to build with artificial intelligence.
Key Facts
Trusted By
ESG-X, Seleon, SolarShare, and 200+ SMEs
Backed By
Supported by leading investors and accelerators
The Team
Daniel Quiroga
Co-Founder & CEO
Daniel combines legal expertise with deep tech experience from scaling unicorns like Personio and Gorillas. Alumnus of LMU, CDTM, and Harvard.
Peter Huo
Co-Founder & CTO
Peter is a Computer Science expert with B.A. & M.Eng. from Cornell University. Previously at Legal Information Institute and Cornell Urban Technology.
Anna Kwon
AI Product Engineer
Anna specializes in computational biology and ML, with experience at Novo Nordisk and the Broad Institute. MIT Computer Science graduate.
Why Now?
SMEs are drowning in repetitive tasks but lack the engineering resources to build custom AI solutions. With the maturation of LLMs, we can now bridge this gap. PUNKU.AI provides the missing layer: Simply describe the workflow you want to automate as if you were to describe it to a co-worker and PUNKU.AI will build it for you.
Scientific Breakthroughs enabling PUNKU.AI
1) Breakthrough algorithms (2025)
ADAS (ICLR’25) introduces Automated Design of Agentic Systems: a meta-agent that programs new agents in code; since programming languages are Turing-complete, this lets the system explore prompts, tool use, and workflows as code. Their Meta Agent Search iteratively writes agents, tests them, and grows an archive of discoveries; discovered agents outperform hand-designed ones and transfer across domains/models.
2) Self-improving agents proven in the lab (May ’25)
Darwin Gödel Machine (DGM) is a self-referential system that edits its own code, alternating self-modification with evaluation, while keeping an archive for open-ended exploration. DGM shows hard metric gains on coding: SWE-bench 20.0% → 50.0% and Polyglot 14.2% → 30.7%. Experiments ran with safety precautions (e.g., sandboxing, human oversight).
3) Lab → workplace validation (May ’25)
TheAgentCompany is a self-hosted workplace benchmark with 175 realistic tasks and simulated colleagues; agents must browse, code, use a terminal, and communicate to finish tasks. Today’s best baseline autonomously completes ~30.3% of tasks, showing meaningful autonomy but room to grow on long-horizon work.
4) Deployable tooling (Feb ’25)
AutoAgent provides a fully-automated, zero-code framework that creates agents, tools, and workflows from natural language; it’s positioned as a top open-source general agent on GAIA.
5) Safety + readiness
The research stack’s sandboxing + oversight (DGM) suggests a staged, responsible path from lab environments to production pilots.
“Algorithms (ADAS, DGM) prove self-improvement; benchmarks (TheAgentCompany) show workplace-relevant autonomy; and tooling (AutoAgent) makes it deployable—so it’s time to productize.”
Sources
- Hu, S., Lu, C., & Clune, J. (2025). Automated Design of Agentic Systems (ADAS). In International Conference on Learning Representations (ICLR 2025). arXiv:2408.08435v2.
- Zhang, J., Hu, S., Lu, C., Lange, R., & Clune, J. (2025). Darwin Gödel Machine: Open-Ended Evolution of Self-Improving Agents. arXiv:2505.22954.
- Tang, J., Fan, T., & Huang, C. (2025). AutoAgent: A Fully-Automated and Zero-Code Framework for LLM Agents. arXiv:2502.05957v2.
- Xu, F. F., Song, Y., Li, B., Tang, Y., et al. (2025). TheAgentCompany: Benchmarking LLM Agents on Consequential Real World Tasks. arXiv:2412.14161v2.
