Skip to content
Renaud Yasin
Portfolio & field notes

Applied-AI systems, owned from brief to production and built through Claude.

Twelve-plus years of regulated-enterprise delivery, now pointed at applied-AI: real systems with live users, evaluation harnesses, and a cost budget.

Currently · MSP delivery lead, WM Digital Sustainability
View the work
12+
years, full lifecycle
6
enterprise programs led
24×7
production ownership
Selected work

Apps I've built with AI

A few of the applications I've built to learn more about AI and its capabilities, each tile drawn from what the app actually outputs. All built using Anthropic Claude.

All projects
Live

Chainsight

Calibrated Bitcoin forecasts computed from my own full node, not bought from an API.

  • Machine Learning
  • Bitcoin
  • On-Chain Analytics
  • Next.js
  • Python
Read the story
Prototype

StockHero

Forecasts stock prices with a leakage-safe, confidence-aware ensemble of five ML models.

Read the story
Prototype

Phiremap

Real-time, browser-based IBIS dialogue mapping: arguments, CRM, and agile on one shared canvas.

Read the story
Prototype

RSS Feeder

Self-hosted marketing intelligence from news feeds, powered entirely by local LLMs.

Read the story
Prototype

SEC Filings Parser

Turns raw SEC EDGAR filings into searchable insider, fund, and company profiles, free.

Read the story
A field guide

Adopt Claude at your company

A vendor-neutral kit I assembled for any company making the case for Claude: the executive one-pager, the governance brief, and prompt packs your teams can use on day one.

Open the kitStart with the one-pager
From the blog

Field notes

Build-and-run lessons, AI-assisted delivery, and the occasional teardown.

All posts
8 min read

The cheapest token is the one you never send

Enterprise Claude bills are a design decision, not a rate you negotiate. The biggest lever isn’t a discount at all; it’s declining to put your whole corpus in context in the first place. A knowledge-graph approach like graphify is the sharpest example, and the rest of the stack compounds from there.

  • Claude
  • AI
  • Cost
5 min read

Chainsight: Forecasting Bitcoin On-Chain, Honestly

Owning a genuinely hard problem end to end as a product owner building through Claude: forecasting Bitcoin from my own full node, and making honesty about uncertainty the whole point.

  • Machine Learning
  • Bitcoin
  • On-Chain Analytics
5 min read

Phiremap: rebuilding IBIS dialogue mapping as a real-time collaborative web app

Owning the brief as a product owner building through Claude: taking a single-user desktop tool and deciding, store by store, how to turn it into a real-time collaborative canvas that also covers CRM and agile.

  • CRDT
  • Real-time collaboration
  • React

How I work

How I work

A Houston-based delivery leader who owns applied-AI systems from brief to production, with 12+ years owning the full lifecycle of enterprise software across banking, financial crimes, telecom, and environmental services. On my own time I set the brief for applied-AI systems and build them through Claude, to learn where the technology earns its keep and to practice the harder discipline of proving it works. My north star: outcomes, not output. AI is the new paintbrush; what I bring is the part it can’t, knowing when the output is good enough to ship.

More about meSee the work
  • 01

    Own the outcome, through production

    Prioritize by business impact and risk, and stay accountable past launch into release and 24×7 operations, not just the demo.

  • 02

    Earn the trust the AI can’t

    Retrieval over brute-force context, human-in-the-loop review, and honest measurement of what the model actually gets right.

  • 03

    Where “it mostly works” isn’t enough

    Regulated, high-scale, security- and compliance-sensitive environments, where the bar has to be provable, not vibes.