Selected Work & Open-Source Architecture
pureQTern & project-zero
Zero-dependency 1.58-bit ternary codec in both C and Rust agreeing to the exact bit on a golden SHA-256 hash, paired with a bare-metal C inference server.
8b-is / MLX-QUANT
Custom low-bit ternary quantization and packed matrix-vector kernels eliminating memory-bandwidth bottlenecks for foundation model inference.
Headroom (PR #1418 — kompress)
Per-content-type bias multipliers for domain-aware compression routing in Headroom.
Osaurus & Osiris
Core Architect on Osaurus (native Swift agent runtime) and Osiris (real-time OSINT intelligence engine).
deepseek-go
High-performance, low-allocation Go client and streaming runtime for foundation model APIs.
ZENO (銭 · 禅 · 善)
Cryptographic double-entry Kintsugi SHA-256 ledger and high-precision settlement engine.
Direct Portal, LinkedIn Contacts & 1-Click Outreach
Toru Nishikawa (西川徹)
Co-founder leading PFN's vertically integrated computing stack (MN-Core, PLaMo, Matlantis).
Daisuke Okanohara (岡野原大輔)
Co-founder & CTO pioneering deep learning infrastructure, low-bit inference, and PLaMo LLMs.
PFN Talent Acquisition & Compiler/Systems Team
Technical recruiting team for MN-Core runtime, Optuna/Rustuna, and PLaMo inference.
Hi! I built a live interactive MN-Core systolic array & Rust systems portfolio specifically for Preferred Networks in Otemachi: https://pfn.vaked.dev — showcasing bit-exact C/Rust 1.58-bit ternary codecs (pureQTern) & custom GPU kernels. Would love to connect! — Péter
Dear Preferred Networks (PFN) Engineering & Talent Team, I created a dedicated interactive systems portfolio and bilingual application dossier for Preferred Networks in Otemachi, Tokyo: • Live Interactive Portfolio (MN-Core Systolic Array Simulation): https://pfn.vaked.dev • Formal 履歴書 + 職務経歴書 (PDF): https://pfn.vaked.dev/resume.pdf • Systems Portfolio (PDF): https://pfn.vaked.dev/portfolio.pdf • Full Technical Application Note: https://pfn.vaked.dev/application Why PFN is my ideal technical home in Tokyo: • Bit-Exact C/Rust Low-Bit Runtimes: I built pureQTern and project-zero — dependency-free 1.58-bit ternary codecs in C and Rust verified to the bit by golden SHA-256 hashes, plus custom Metal GPU kernels (MLX-QUANT) achieving 12.80× compression. • High-Throughput Systems & Open Source: Core Architect on Osaurus (osaurus-ai/osaurus) & Osiris (simplifaisoul/osiris), contributor to Headroom (PR #1418), and 9+ years leading high-concurrency Go/Rust/Python systems (ASAPP, DailyWire). Warm regards, Péter József Lódri https://www.linkedin.com/in/peter-lodri/
Technical Stack & Capabilities
Rust, C, Swift 6, Go — SIMD, custom Metal/WebGL2 GPU kernels, BitNet b1.58 {-1, 0, +1} ternary quantization (12.80× memory reduction), zero-dependency binaries.
Unity, Three.js, WebGL2, GLSL/HLSL shaders, oscilloscope vector synthesis, LiDAR/geospatial 3D telemetry (Osiris), procedural Web Audio.
Core Architect on Osaurus (osaurus-ai/osaurus) & Osiris (simplifaisoul/osiris), contributor to Headroom (PR #1418 kompress router), creator of AgentField verifiable DAGs.
9+ years shipping high-concurrency Go/Python/Rust/TS systems (ASAPP, DailyWire 110k concurrent users, institutional fintech ETL). Fluent English, native Hungarian, conversational multi-lingual + daily Japanese in Tokyo.
Why Preferred Networks (PFN · プリファードネットワークス)
Preferred Networks stands virtually alone in Japan in owning the entire vertical stack — from custom MN-Core systolic matrix silicon and compiler toolchains down to sovereign Japanese foundation models (PLaMo) and black-box optimization (Optuna / Rustuna). My engineering obsession is that exact boundary where mathematical structure meets hardware memory layout: writing zero-dependency C and Rust ternary codecs (pureQTern) that agree to the single bit, and custom GPU kernels (MLX-QUANT) that cut memory bandwidth by 12.80×.
Preferred Networks（PFN）が推進する独自AI半導体「MN-Core」シリーズから国産基盤モデル「PLaMo」、最適化基盤「Optuna / Rustuna」に至る垂直統合技術は、世界のAI計算基盤の中でも他に類を見ないものです。C言語とRustで1ビットの狂いもなく一致する三値量子化コーデック（pureQTern）やカスタムGPUカーネル（MLX-QUANT）をゼロ依存で実装してきた私のシステム工学スキルを、大手町のPFNチームでMN-CoreランタイムおよびPLaMo推論基盤の高速化に捧げたいと強く志望しています。