LLM Consulting & Engineering
A full-stack engineer fluent in business — I support clients from LLM / SLM development and adoption to IPO readiness and engineering organizations.
Since 2008 I have built systems across industries — healthcare, finance, payments, education, gaming, hardware — and across every layer of the stack. More recently I have been shipping working LLM applications, such as character role-play and medical document checking systems. As CTO, I also led a company through its IPO on the Tokyo Stock Exchange Growth Market. Building on that breadth, I support clients from PoC to production operations and team building.
More than any single skill, my greatest strength is the breadth of experience spanning both technology and business domains. LLMs only deliver real value when they reach deep into actual operations. Being able to discuss your industry context and the realities of your existing systems from the very first meeting makes proposals faster and more precise.
From fine-tuning (SFT, LoRA, continued pre-training) to RAG, agent design, and building evaluation and operations infrastructure. I have delivered working systems including character role-play, medical document checking, and STT / TTS. I support projects with production operation in mind — not stopping at the PoC stage.
As Executive Officer & CTO of pluszero, Inc., I was responsible for the technology side of its 2022 IPO on the Tokyo Stock Exchange Growth Market (ticker 5132), leading the build-out of an engineering organization, security, and information governance that could withstand an IPO. I can now support the technology side of companies aiming for their own listing, as someone who has done it.
Across healthcare, finance, payments, e-commerce, education, gaming, and more, I have joined projects from the planning and requirements phase and supported new business launches. I can frame technology investment decisions in terms of business value and translate between management and engineering teams.
From frontend, server-side, and infrastructure (AWS) to mobile and embedded, I have built systems hands-on at every layer. I can design realistic, pragmatic ways to integrate LLMs into existing systems.
A selection of AI products I have built hands-on. My style is to turn an idea into something working as fast as possible and validate it. Demo videos and write-ups for each product are coming soon.
A testbed for how cheap and fast an LLM service can run on a small model with CPU-only inference: a 1.2B SLM distilled and DPO-tuned from a Gemma 31B teacher, served on AWS Lambda by a hand-built Rust CPU inference engine. The vehicle is a dream-reading web app that returns symbol meanings and a gentle reading. A live demo is available.
A web app that turns browser microphone input into live Japanese-English captions and interpretation, displayed side by side in fixed English and Japanese panes. Silero VAD detects utterance boundaries, and Groq's low-latency inference (Whisper + Llama 3.3 70B) handles recognition and translation. A two-stage design — provisional captions on short silences, commit-and-translate on longer ones — balances fast source-text display with stable translations.
A web app for practicing how you speak — intonation, pauses, emphasis, and pace — in Japanese and English. Record a short free-form utterance and it analyzes monotony, hesitations, filler words, and more, visualizes them as word-level overlays alongside a pitch curve, and an LLM gives concrete feedback on exactly what to fix and how.
Validating "Efficient Thinking" on Qwen 3.5 9B / 3.6 27B: Self Distill selects only the good reasoning traces the model itself generates and distills from them, reducing reasoning tokens while preserving accuracy. A practical example of making SLMs viable by cutting inference cost and latency.
Advisory for companies on the road to IPO, from someone who has been through it. Building an engineering organization, security, and information governance that withstand listing scrutiny; supporting audits and reviews; and helping executive teams make technology investment decisions — including development organizations for the LLM era.
Hands-on support for fine-tuning, RAG and agent development, and the design and implementation of evaluation infrastructure — including knowledge transfer to your team.
I join new businesses and AI products as a partner on the business side, not as a contractor. Structures including revenue share and equity are open to discussion. We start by building small and validating together.
One-off technical consultations, second opinions, technical due diligence, and more. Available from a single hour — a good way to start the conversation.
Committing to an hourly rate or a monthly retainer with someone you have just met is a big ask. My recommendation is the success fee — I take on the risk.
| Model | Delivery risk | Business-outcome risk |
|---|---|---|
| Retainer (monthly / hourly) | Client | Client |
| Success fee | Me | Client |
| Revenue share / equity | Me | Shared |
Delivery risk is whether the agreed deliverable gets built; business-outcome risk is whether it produces business results (revenue and profit). Under the success fee I carry the risk through to the deliverable's completion, while the later business outcome stays the client's risk. With revenue share / equity, that business-outcome risk is shared too.
My focus is generative AI and LLMs / SLMs with the surrounding software development, plus engineering organizations and IPO readiness. Rates are quoted individually based on scope, duration, and engagement style.
A selection from the many projects I have worked on since 2008.
For inquiries and engagements, feel free to reach out by email or via DM on X — even if you just want to talk things through first.