🜂 sakana.vaked.devSakana AI (サカナAI) · Péter József Lódri

Sakana AI (サカナAI) — Complete Application Dossier & LinkedIn Kit

Applicant: Péter József Lódri (LODRI PETER JOZSEF · ロドリ・ペーテル・ヨージェフ)
Target Role: AI Systems & Multi-GPU Infrastructure Engineer (Low-Bit Inference & Agentic Swarms)
Location: Minato-ku, Tokyo (港区・虎ノ門 / 赤坂)
Compensation Band: ¥15,000,000 – ¥28,000,000+ JPY + Equity
Dedicated Live Subdomain: https://sakana.vaked.dev
Formal Résumé PDF: https://sakana.vaked.dev/resume.pdf
Engineering Portfolio PDF: https://sakana.vaked.dev/portfolio.pdf
LinkedIn Profile: https://www.linkedin.com/in/peter-lodri/
GitHub & OSS Architect: github.com/peterlodri-sec · github.com/8b-is · osaurus-ai/osaurus · simplifaisoul/osiris · headroomlabs-ai/headroom#1418

Direct Application & LinkedIn Contact Directory

Official Careers Portal: https://sakana.ai/careers/
Direct Recruiting Email: careers@sakana.ai
LinkedIn Company Page: https://www.linkedin.com/company/sakana-ai/
LinkedIn People Directory: https://www.linkedin.com/company/sakana-ai/people/
LinkedIn Recruiter & Team Search: Find Sakana AI (サカナAI) Recruiters & Engineers

Key Leadership & Talent Contacts

David Ha (Co-Founder & CEO, Sakana AI (Former Head of Google Brain Tokyo)): https://www.linkedin.com/search/results/people/?keywords=David%20Ha%20Sakana%20AI — Co-founder leading evolutionary model merging, AI Scientist & collective intelligence in Tokyo.
Llion Jones (Co-Founder & CTO, Sakana AI (Co-Author of 'Attention Is All You Need')): https://www.linkedin.com/search/results/people/?keywords=Llion%20Jones%20Sakana%20AI — Co-founder & CTO overseeing novel foundation architectures and inference systems.
Ren Ito (伊藤錬) (Co-Founder & COO, Sakana AI): https://www.linkedin.com/search/results/people/?keywords=Ren%20Ito%20Sakana%20AI — Co-founder leading operations, strategic partnerships, and Tokyo talent expansion.

LinkedIn Connection Note (< 300 chars)

Hi! I built a live interactive Boids + 1.58-bit ternary inference portfolio specifically for Sakana AI's Systems & GPU Infrastructure team in Tokyo: https://sakana.vaked.dev (12.8x BitNet compression + Osaurus/Headroom OSS architect). Would love to connect! — Péter

LinkedIn InMail / Direct Email Body

Dear Sakana AI Founding & Engineering Team,

I built a dedicated interactive showcase and bilingual engineering dossier for Sakana AI's AI Systems & Multi-GPU Infrastructure Engineer role in Tokyo:

• Live Interactive Portfolio (Emergent Boids + Ternary Lattice): https://sakana.vaked.dev
• Tailored Resume / 職務経歴書 (PDF): https://sakana.vaked.dev/resume.pdf
• Systems Portfolio (PDF): https://sakana.vaked.dev/portfolio.pdf
• Full Application & Architecture Note: https://sakana.vaked.dev/application

My work centers on two pillars at the heart of Sakana AI's mission:
1. Low-Bit Foundation Model Compression (8b-is/MLX-QUANT & pureQTern): 12.80× memory reduction via BitNet b1.58 {-1, 0, +1} ternary quantization with custom Metal GPU kernels and zero-dependency C/Rust codecs.
2. Collective Multi-Agent Systems: Core Architect on Osaurus (osaurus-ai/osaurus — native Swift offline agent harness), contributor to Headroom (headroomlabs-ai/headroom PR #1418 — kompress domain-routing bias multipliers), and builder of AgentField multi-reasoner DAG orchestration.

Warm regards,
Péter József Lódri
https://www.linkedin.com/in/peter-lodri/

02 — Sakana AI (Tokyo, Minato-ku)

Role: AI Systems & Multi-GPU Infrastructure Engineer (Evolutionary Model Merging, Low-Bit Inference & Agentic Systems)

Estimated Compensation: ¥15,000,000 – ¥28,000,000+ JPY + Equity (Tokyo Hybrid / On-Site)

Application Portal: https://sakana.ai/careers/ · Direct Email: careers@sakana.ai

Resume to Attach: [../teamlab/lodri-resume.pdf](../teamlab/lodri-resume.pdf) & [../resumes/cv-peter-lodri-taiko.pdf](../resumes/cv-peter-lodri-taiko.pdf)

1. ELI5: What Does Sakana AI Do & Why Do They Need You?

What they do: Founded in Tokyo by former Google Brain researchers (David Ha & Llion Jones, co-author of the original Attention Is All You Need Transformer paper), Sakana AI builds nature-inspired AI: instead of training one giant, expensive model, they evolve swarms of smaller models (Evolutionary Model Merging) and autonomous AI Scientists (The AI Scientist).
Why you are a SOTA match: Sakana AI needs systems engineers who can make swarms of models run ultra-fast on GPUs and Apple Silicon/edge hardware with verifiable determinism. You have already shipped:

8b-is/MLX-QUANT & pureQTern: 1.58-bit ternary (BitNet b1.58 {-1, 0, +1}) quantization and custom Metal GPU kernels (12.80× compression over FP16).
8b-is/cogito (v0.4.0): Zero-float-drift Pq arithmetic, hardware CLZ-seeded isqrt_pq (1,000,000 calls in 5.52 ms), and BFP-64 64-trit POPCNT bitplanes.
Multi-Reasoner Agentic Infrastructure: Production AgentField multi-reasoner pipelines and verifiable memory ledgers (mem-16-10).

2. Ready-to-Send Application Email / Cover Note (`careers@sakana.ai`)

Subject: Application: AI Systems & Low-Bit Inference Engineer — Péter József Lódri (Rust / Metal & GPU Kernels / Multi-Agent Systems, Tokyo)

Dear Sakana AI Hiring Team,

I am a Tokyo-based Principal Systems Engineer & Independent Researcher (8+ years across ASAPP, Plandek, Virality Data, and DailyWire) applying for AI Systems & Infrastructure Engineering at Sakana AI.

Sakana AI's thesis—that nature-inspired collective intelligence, evolutionary model composition, and efficient specialized models outperform monolithic brute-force scale—is the exact architectural principle behind my open-source systems work:

1. Low-Bit Ternary Quantization & Custom GPU Kernels (`MLX-QUANT` & `pureQTern`):
   - Built zero-dependency C and Rust inference engines and Apple Silicon Metal GPU kernels for BitNet b1.58 ({-1, 0, +1}) ternary models, achieving 12.80x weight compression over FP16 with bit-exact golden-hash parity between CPU and GPU.
2. Deterministic Sub-Millisecond Kernels (`8b-is/cogito`):
   - Engineered BFP-64 (64-trit POPCNT bitplanes with adversarial bit-flip sanitization) and zero-float-drift rational lattices (`Pq`) executing 1,000,000 integer square-root evaluations in 5.52 ms.
3. High-Throughput Distributed & Agentic Systems:
   - At ASAPP, architected Go/Kotlin/Kubernetes microservices that cut Fortune-100 AI customer-agent latency by 50–70%.
   - Built `mem-16-10` (`#![no_std]`, zero-alloc Rust verifiable memory ledger) and self-host a 30+ service NixOS/nix-darwin compute fleet.

Code & Live Systems:
- GitHub: https://github.com/peterlodri-sec & https://github.com/8b-is
- Interactive Technical Portfolio: https://teamlab.vaked.dev
- Verifiable Compute Engine: https://github.com/8b-is/cogito

I am currently in Tokyo, hold an EU passport (Hungary, BJ3429501), and would love to discuss how my Rust, GPU kernel, and multi-agent infrastructure work can accelerate Sakana AI's research and enterprise deployments.

Best regards,
Péter József Lódri
peter.lodri@gmail.com | +36 20 391 5161
https://github.com/peterlodri-sec