Jimi Vaubien

01 — The Problem

Your team is doing work that AI should be doing.

01

Workflows eating 10–30 hours a week. Manual processes nobody has questioned.

02

AI implementations that work in demos but break under real data. Never deployed.

03

7 years shipping AI in regulated finance. Systems that stay in production for years.

Lead generation Document processing Email triage Data extraction Report generation Content Creation

02 — The Offer

One workflow.
Automated.
Guaranteed.

Pick your most painful manual workflow. In 2–4 weeks I build the AI system. Clear metrics upfront. You own the result.

Pricing

Book a call →

Let's discuss your workflow and budget

Miss the agreed metric? You don't pay the final milestone.

Week 1

Workflow Audit

Map the process. Define success. Agree on "done".

Week 2–3

Build

Agents, pipelines, integrations. The real work.

Week 4

Handoff

Testing, training, documentation. Your team owns it.

Book a free 20-min call →

Taking 2 pilot clients. Pilot pricing for first 3 engagements.

03 — Track Record

Production AI. Not demos.

25%

Workday Returned

Financial Document Processing

A financial firm's team was manually reading emails, extracting data, and entering it into audit software all day. We automated the entire pipeline: any document type, straight into their system."

Azure OpenAI Graph API Python

90%

Recall Rate

Enterprise RAG System

A company had years of documents and data sitting unused. Their team couldn't find answers fast enough. We built an AI system that finds the right information instantly across millions of documents and databases simultaneously.

Azure AI Search Python SQL LLMs

3h

Saved / Agent / Day

High-Throughput LLM Service

A company was drowning in daily support tickets. Their team spent all day copy-pasting the same answers. We automated 68% of tickets completely: response time dropped from 4 hours to 45 seconds.

vLLM Llama 3 Azure A100 Python

04 — Fit Check

This works if you're:

A founder or ops lead at a 10–100 person company

Watching your team repeat the same tasks weekly

Tried no-code tools and hit their limits

Need AI but don’t have an ML engineer internally

Want production-grade, not a demo

Not for you if:

You need someone to manage existing AI tools

You want a strategy deck with no implementation

You’re looking for the cheapest option

05 — About

CV · GitHub · YouTube · LinkedIn

Lead AI Engineer · Consultant

Started in AI research at EPFL and NEC Labs Princeton. 5 years shipping production ML at Visium Technologies — banking, pharma, e-commerce. Then founding engineer at Prometee, building a GenAI platform now running in 4 enterprise financial clients.

Not a consultant who learned AI last year. An engineer who's been shipping this since before it was called GenAI.

Fraud Detection

2+ yrs in private bank

Recommenders

20% revenue uplift

RAG Pipelines

Multi-TB corpora

AI Agents

100s of txns/day

06 — Open Source

RustGPT

699 ★

ChatGPT UI in Rust + HTMX. Zero JS framework. Featured on HN.

FuseAI

293 ★

Self-hosted ai chat web app. Multi-user. Docker deploy.

Kiru

8 ★

Text chunking for RAG. Rust core, Python bindings.

LazyCodr

45 ★

AI CLI for PR descriptions, commits, and docs.

07 — Talks

2024 Website to RAG: building production search with vector indexing AI Tinkerers Lausanne

2023 Building Edge-Ready Web Applications on Cloudflare Webmardi #136

2023 Building Web Apps with HTMX LauzHack Workshop

Let's talk about
your workflow

20 minutes. No pitch. No proposal until we both agree it makes sense.

[email protected]

I turn expensivemanual workflowsinto AI systems thatrun themselves.