Mawu  ·  Independent AI Practice

Production AI for places where the rules are strict and the margin for error is thin.

More than a decade building agentic and machine-learning systems — most recently at national scale in healthcare. Available for architecture, implementation, and advisory work with teams in regulated industries and the public sector.

Capabilities

Agentic system design

Multi-agent architectures, tool and API integration, orchestration, and LLM routing — with the evaluation harness that keeps them reliable once they are carrying real work.

Retrieval & LLM infrastructure

RAG pipelines, embeddings and reranking, model serving and inference, and the vector and graph data layers underneath them.

Production hardening & MLOps

Taking a prototype the rest of the way: deployment, observability, cost and latency control, and continuous delivery for models.

Governance for regulated settings

Tenant isolation, access controls, audit trails, and responsible-adoption practices that hold up under HIPAA, CMS, and public-sector constraints.

AI adoption & context setup

Connect Claude to how your team already works — your tickets, your repos, your infrastructure, your roadmap — so it operates with real context from day one, then build the habit of working with it from there.

Selected work

CaseOS

A production multi-agent AI platform

Built and operated end to end for legal case work: agent orchestration, self-hosted model serving behind a routing layer, retrieval infrastructure, and a graph-based entity-resolution layer — multi-tenant, with strict data isolation between clients.

Healthcare payer

Enterprise document intelligence

Architected an AI document-intelligence platform on AWS for a national managed-care organization. Cut turnaround on a core document workflow from roughly three days to under five minutes.

Public sector

Government middleware & integration

More than a decade delivering mission-critical middleware and integration platforms for the State of Missouri and the City of St. Louis under formal statements of work, at the uptime and security bar public operations require.

Via Slalom

Fortune 50 healthcare delivery

Cloud, DevOps, and ML-infrastructure engagements embedded inside Fortune 50 healthcare clients, from discovery through production, 2018–2021.

Engagement models

Project

Scoped build

An architecture or implementation engagement under a statement of work. Fixed outcome, defined timeline.

Fractional

Embedded architect

Ongoing part-time capacity as your AI architect — a set number of days each month, inside your team.

Advisory

Decision support

A retainer for architecture review, technical strategy, and second opinions — without the day-to-day build.

Who you're working with

Pierre Kpodar founded Mawu in 2015. Through 2026 he led machine-learning and cloud engineering at a national managed-care organization, owning production LLM systems in a HIPAA- and CMS-governed environment. Before that: senior consultant at Slalom, and technical lead on edge computer-vision systems for utility infrastructure.

AWS Certified Solutions Architect. Based in St. Louis; works remotely and on-site. English and French.

Get in touch

Tell me about the system you're trying to build, or the decision you're trying to make. I'll tell you plainly whether I can help.

pierre@mawu.io