Logo Nathan.R
Profile Picture

Who Am I?

Nathan RIHET (ネイサン・リヘ) - Software Engineer & Visual Storyteller ready to build software that matters.

How I Got Here

I never planned to become a software engineer. Started with a camera in Nice, studying networks because it seemed practical. But somewhere between debugging my first Python script and shipping my first React app, I fell in love with building things that matter.

My Master's in AI wasn't about chasing trends. I wanted to understand how machines could actually help people find what they need. That obsession led me to RAG systems, where I learned that good search isn't about keywords anymore. It's about understanding intent.

March 2025: I quit my job at Capgemini, packed two suitcases, and flew to Osaka. No job lined up. Just a conviction that Japan's tech scene needed builders who think differently. Three months later, I've launched a SaaS with paying customers and I'm teaching myself Japanese between debugging sessions.

What I've Built (And Why It Matters)

Capgemini GenAI Division (2024-2025)

They hired me to build "AI stuff." What I actually did was fix how 200 people worked every day.

The problem: Business analysts spending 15 minutes searching through outdated wikis for salary calculation rules. My solution: A RAG system using MongoDB Atlas Vector Search that understands questions, not just keywords. Search time dropped to 2 seconds. But the real win? People stopped complaining about documentation.

Also led our Flask to FastAPI migration. Not because FastAPI is trendy, but because async Python actually matters when you're running inference on LLMs. Response times dropped 60%, but more importantly, we could finally handle concurrent requests without the server crying.

Terraform came later. Got tired of deployments taking my entire afternoon. Wrote infrastructure as code, automated everything. Four hours became 30 minutes. Now I could actually go home on time.

Building My Own Thing (2025-Now)

DocsRetriever started as a side project while at Capgemini. Launched it properly after moving to Japan. Currently 10 paying customers who trust me with their document search needs.

Technical choices that matter:

  • MongoDB Atlas because vector search just works
  • FastAPI because I need speed and good docs
  • Custom VPS deployment because I like controlling my stack
  • Docker orchestration I built myself (probably should use K8s i know, but maybe next time i would use it with your team?)

Also building something for the Japanese market. Can't share details yet, but it involves understanding how Japanese users think about task management differently.

Teaching What I Know

Université Côte d'Azur asked me to teach their AI Master's students. Seven workshops on building real LLM applications. No theory, just "here's how you actually ship this stuff."

Best moment: A student messaged me last week. She got hired at a startup using the RAG system we built in class. That's why I teach.

My Tech Stack (And Why I Chose It)

TypeScript: Because JavaScript with types is just better. Three years of production React taught me that runtime errors in production aren't fun. I'd rather fight with the compiler than debug customer complaints.

Python: Still the best for AI work. FastAPI for backends, LangChain for LLM orchestration. The ecosystem is unmatched. Plus, async Python finally makes sense.

React/Next.js: I've tried Angular, Vue, Svelte. Keep coming back to React. Next.js App Router changed how I think about server/client boundaries. Shipping RSCs to production was a game changer.

The AI Stack:

  • LangChain for orchestration (though I'm starting to build more custom chains)
  • MongoDB Atlas Vector Search (tried Pinecone, too expensive for my use case)
  • OpenAI APIs, though experimenting with open models
  • Custom evaluation frameworks because "it seems to work" isn't good enough

Infrastructure: Docker everything. GitHub Actions for CI/CD. Nginx reverse proxy. Ubuntu VPS because I like SSH-ing into my servers. Call me old school but my first diploma was in networks, so I still enjoy the command line.

Photography Changed How I Code

Six years of photography taught me things bootcamps don't:

Composition matters in code too. A well-structured component is like a well-framed shot. Everything has its place, nothing is accidental.

Light reveals form. In photography and in debugging. Sometimes you need to shine light from different angles to see the bug.

The decisive moment exists in shipping too. There's a point where more tweaking makes things worse, not better. Ship it.

My photography work spans three countries now. Each place taught me to see differently. Quebec's harsh winters, Provence's soft light, Osaka's neon chaos. Each influenced how I approach problems.

Why Japan?

People ask why I left a comfortable job in France for uncertainty in Japan. Simple: I was getting too comfortable.

Japan's approach to craft resonates with how I see software. The attention to detail. The respect for process. The balance of tradition and innovation. Plus, debugging code at 2 AM hits different when you can walk to a konbini for coffee.

Currently grinding through Japanese lessons. N5 level isn't impressive, but I'm committed. Already having basic conversations with local developers. Language barriers force clearer thinking about architecture. Win-win.

What I'm Looking For

Not just any job. I want to build products people actually use. Teams that ship regularly. Problems worth solving.

Specifically:

  • Full-stack or backend roles where I can leverage my AI experience
  • Companies that value both technical excellence and user experience
  • Teams building for Japanese or global markets
  • Visa sponsorship (obviously)

I bring more than code. I bring experience shipping to production, teaching complex concepts, and building from zero to revenue. Plus, I can document your architecture and make it look good (photographer perks).

Current Projects

DocsRetriever: Growing slowly but sustainably. Each customer teaches me something new about enterprise search needs.

Japanese Market App: Under wraps for now. Solving a real problem I discovered living here.

Open Source: Contributing to LangChain docs. Found too many errors when I was learning. Fixing them for the next person.

Photography: Documenting Osaka's hidden corners. Keeps me sane between coding sessions.

Let's Talk

I'm in Osaka, available immediately, and ready to contribute. Whether you need someone to architect your next AI system, optimize your Python backends, or just want to debate tabs vs spaces over okonomiyaki, reach out.


Find Me

📧 nathan.rihet06@gmail.com 📑 Curriculum Vitae 💼 LinkedIn
🐙 GitHub
📷 Photography
📍 Osaka, Japan (but trains here are amazing, so distance isn't a problem)

P.S. - If you're reading this and thinking "this guy seems alright," you're probably right. If you're thinking "too casual for our corporate culture," we probably wouldn't work well together anyway. Life's too short for boring tech.