Platform leadership in practice.

The platform, and the systems around it.

I lead Valeo Financial Advisors' platform and data engineering team and partner with executive, finance, and investment leaders. Our scope covers the software, data, and application cloud resources behind internal products.

Flagship platform

Production · Technical lead & platform owner

VAMOS — the firm's internal advisor platform

VAMOS replaced the spreadsheet behind practice composition and compensation. I lead platform work with a team of contributors and partner with finance leadership.

Impact: Replaced a spreadsheet behind a roughly 200-person quarterly process
Delivery: Inherited prototype to production in about one month
Leadership: Architecture, delivery standards, review, and production accountability

React
TypeScript
FastAPI
PostgreSQL
Azure

Read the writeup →

Advisor and data tools

In development · Technical lead

PRISM — investment information for advisors

I’m developing an advisor-support tool in collaboration with investment leadership and Compliance. It brings structured, cited information into the advisor’s workflow while keeping recommendations and client advice with the advisor. Deployment preparation is underway.

Partners: Investment leadership and Compliance
Workflow: Structured information with citations
Boundary: Recommendations and client advice remain with the advisor

React
TypeScript
FastAPI

Deployed MVP · limited use · Designer & builder

Fabric-MCP — helping leaders explore firm data

I’m building a way for executive leadership to explore firm data through natural-language questions, with controlled, read-only access to the data platform. An initial MVP is deployed in limited use. I’m now developing the context layer: business definitions, firm terminology, and caveats that help the assistant interpret the data appropriately.

Access: Controlled, read-only connection to the data platform
Context: Business definitions, firm terminology, and caveats are being curated

Python
MCP
Microsoft Fabric
T-SQL

Application delivery

In internal use · Platform architect & owner

Armada — a deployment platform for AI-built apps

Armada hosts internal applications used by colleagues. Reliable deployment and recovery work continues as the self-service workflow develops.

Purpose: Self-service path for internally built applications
Boundary: Gateway SSO, isolated containers, and deny-by-default egress
Governance: Ownership, access, infrastructure, and support boundaries defined at the platform level

Python
FastAPI
React
PostgreSQL
Azure

Read the writeup →

Implementation lead (historical)

Salesforce — original implementation

I led Valeo's original Salesforce implementation, from requirements and data model through configuration, development, and the Phase 1 rollout.

Delivery: Led requirements, data model, configuration, development, and the Phase 1 rollout

Salesforce
Flow
SOQL

How the team builds

Production · Engineering systems designer

CARL — an autonomous software-development pipeline

CARL drives agents from ticket to pull request through isolated workspaces, model routing, and human approval gates. The full autonomy path did not outperform an engineer directing good tools, but its Review Layer now gives every team pull request a consistent automated review.

Current use: Reviews every VAMOS pull request into dev
Finding: Caught a browser-only timezone defect after CI was green
Ownership: Designed and built the pipeline and review layer

Review evidence

VAMOS PR #283

“Adoption chart month labels are off-by-one in Eastern time”

The underlying data was correct, but every chart label was wrong for Eastern-time users. CI stayed green because its runner used UTC.

Fixed in the follow-up commit, with regression tests added.

Real finding from a VAMOS pull request, lightly edited for public display.

Python
Claude Agent SDK
GitHub API
systemd

Read the writeup →

In production · Technical standards & enablement

VFA-Tools & VFA-Context — a portable agent harness

Shared engineering workflows and domain knowledge shaped by delivery, review, and incident lessons, with practices used across repositories.

Distribution: Shared workflows and context embedded in repository work
Enforcement: Pull requests are reviewed against the exact submitted commits
Layers: Workflow skills, domain context, and deterministic hooks

Embedded engineering harness

.claude/
├── hooks/
│   ├── pr-guard.mjs
│   └── submodule-integrity.mjs
├── rules/
│   ├── backend-python.md
│   └── security-gates.md
└── skills/ → VFA-Tools
    ├── VFA-Context/
    ├── vfa-pre-pr/
    ├── vfa-review-code/
    └── vfa-review-data-model/

Session start verifies that the versioned tools and context are present.
Editing loads the rules and domain context relevant to the work.
A pull request into dev requires a passing review on the exact commits being submitted.
Any later commit invalidates that pass and requires another review.

Claude Code skills
Markdown
git submodules
Conventional Commits

Early development · Team leadership & collaboration

VFA-Foundation — shared platform packages

An early, team-led effort to develop shared UI and authentication foundations for reuse across internal applications.

TypeScript
React
Python

Operational leverage

Designer & builder

Operational Automation — eliminating the firm's busywork

A set of unattended workflows that remove recurring document and data chores from the firm's operating rhythm.

Outcome: Turns recurring document work into unattended workflows
Scope: Secure file intake, PDF splitting, and Salesforce automation
Design goal: Runs without ongoing operator intervention

Python
systemd
Salesforce Flow