Agentic AI Architect | Manager & Solution Architect at Ashling Partners | UiPath MVP
Open-source contributor to Microsoft Agent Framework and Microsoft Agent Governance Toolkit.
I design enterprise-grade agentic systems that move AI agents from demos into reliable business workflows, with a focus on orchestration, context engineering, governance, security, human oversight, and measurable operational value.
My current technical interests include agentic automation, WebMCP, agent reliability, context management, tool calling, workflow orchestration, and enterprise AI architecture.
Code, fixes, and samples contributed to Microsoft's open-source platforms for building and governing AI agents: 4 merged pull requests, with 4 more in review (September 2026).
Microsoft's framework for building, orchestrating, and deploying AI agents and multi-agent workflows in Python and .NET.
| Status | Pull request | What it changed |
|---|---|---|
| Merged | #8231 Wake workflow streaming on iteration completion | Replaced an up-to-50 ms polling wait between supersteps in the Python workflow runner with an immediate completion signal, so fast executors are no longer throttled. |
| In review | #8232 Container-hosted workflow sample (.NET) | Runnable ASP.NET Core workflow hosted in a non-root Linux container, verifiable end to end without model credentials. |
| In review | #8437 Clarify Harness client API selection (.NET) | Documents how the Harness client selects Responses or Chat Completions before users configure reasoning and tools. |
| In review | #8438 Cumulative Harness console session tokens (.NET) | Shows total token usage per agent session alongside the latest-call counts. |
Policy enforcement, zero-trust identity, execution sandboxing, and reliability engineering for autonomous AI agents.
| Status | Pull request | What it changed |
|---|---|---|
| Merged | #3909 Bind sidecar decisions to policy-load provenance | Every governance decision now identifies the exact policy set that was loaded, including files that failed to load, through a content-addressed manifest. |
| Merged | #3572 Preserve audit verification across rollover | Keeps the Claude Code hash-chained audit log verifiable after rollover; previously verification broke permanently and the fail-closed path denied every later decision. |
| Merged | #3573 Scope relay inbox deduplication per recipient | Closed a bug that let one sender suppress offline message delivery to a different recipient. |
| In review | #3910 Preserve native execution identity in governance evidence | Carries native Agent Framework call and session IDs into hash-covered governance audit records. |
- W3C TPAC 2026, AI & Society proposal: Who Is the User When an AI Agent Uses the Web?
- WebMCP contribution: Clarify observation context management
- Exploring how emerging Web standards can preserve user intent, interoperability, and meaningful human control as AI agents increasingly act on the Web.
Private, local PDF redaction driven by plain-English instructions, designed around deterministic processing, auditable behavior, and safe handling of sensitive documents.
A developer-focused workflow for scaffolding, validating, deploying, and verifying UiPath Coded Web Apps and Coded Action Apps.
A coded agent for tariff and trade-policy validation in enterprise supply-chain risk workflows.
- Agentic AI architecture
- Enterprise AI agents
- Context engineering
- Agent reliability and evaluation
- Tool and function calling
- Human-in-the-loop systems
- Workflow orchestration
- AI governance and security
- UiPath agentic automation
- WebMCP and the agentic Web
- UiPath MVP, 2024
- UiPath MVP, 2025
- UiPath MVP, 2026
- UiPath Coded Agents Hackathon, Top 7 finalist
- How should enterprise agents preserve user intent across long-running workflows?
- What should be deterministic versus model-driven in production agent systems?
- How should agent memory and context be managed without creating hidden reliability failures?
- How can Web standards enable agent autonomy without reducing human control?
- How do we make agent behavior observable, testable, and governable?
Open-source contributions and technical opinions are my own and do not necessarily represent Ashling Partners.