I am a forward deployed engineer. I solve complex business problems from the inside: I take the problem from the people who have it, find where it actually breaks, and ship the fix into the systems they already run, most often Microsoft 365 and Azure.
One contract, more than 4,700 hours so far, as a software and automation engineer for a US managed service provider on US Pacific hours. My time there is logged against 45 client organisations and 22 projects.
What I have built there:
• An internal automation platform on Azure that I proposed, built, hardened and still patch: n8n in queue mode, now the production home of five systems.
• Moved the AI platform's database to Entra ID sign-in with password login off, carrying 22 workflows over with no workflow edits.
• AI ticket intelligence over the live PSA, extracting a structured record from 2.1 million ticket events and calling the model only where a ticket actually changed.
• A customer-sentiment escalation loop into Teams, answered 108 times, so an unhappy customer reaches a manager early instead of days later.
• A tool-calling AI assistant in Teams that answers staff questions from live systems of record through twelve tool workflows, 126 sessions over 269 days in production.
• A 26-job monitoring platform across 55 Microsoft 365 and Azure tenants, declared in four Bicep files, replacing the helpdesk's manual per-tenant checking.
• Tenant tooling that holds no client secret: managed identity to Key Vault to a per-tenant Microsoft Graph token exchange, with monthly rotation across 54 customer tenants that proves each new credential before deleting the old one.
• Five internal Power Apps rebuilt as React applications on Azure Static Web Apps with a Functions API.
• A ConnectWise MCP server fork with per-member impersonation, Key Vault storage and per-call auditing, deployed as the Claude.ai organisation connector on its second day.
• Led an acquired company's Jira migration into a 2,000-seat tenant: 83,722 issues in one cutover.
The stack I work in:
Microsoft 365: Entra ID, Intune, Autopilot, Exchange Online, SharePoint, Teams, Defender, Microsoft Graph, Conditional Access, PIM, GDAP, Partner Center, Lighthouse.
Azure: Functions, Container Apps, App Service, Static Web Apps, Key Vault, Automation, Azure SQL, Blob Storage, Event Grid, API Management, AI Search, Application Insights, Virtual Desktop, FSLogix, Nerdio, Bicep.
Power Platform: Power Automate, Power Apps, Power Fx, Dataverse, custom connectors, Power BI.
AI: Azure OpenAI, Claude, Claude Code, Model Context Protocol, pgvector, Cohere, Retell AI, Bot Framework.
DevOps: Azure DevOps (Boards, Repos, Pipelines), GitHub Actions, Docker, Linux.
Code: Python, TypeScript, JavaScript, PowerShell, C#, SQL, Bash, React, Next.js, Node.js, Tailwind CSS, React Native, Expo, Electron, .NET.
Data: SQL Server, PostgreSQL, MySQL, Redis.
Integrations: ConnectWise PSA, n8n, Jira, Confluence, Zabbix, REST and GraphQL APIs, webhooks.
Also: AWS, Firebase, Vercel, Appwrite, Raspberry Pi, OpenCV, PyTorch, TensorFlow.
Luis Daniel P. earns an estimated $6.1k/mo. That's 5.1× the typical freelancer and more than 99.67% of everyone we track.
What I am actually good at is inheriting things. A client's decade-old production CRM came to me when its in-house engineer left. I have been its only developer since, and it now has a repository, a deploy path and documentation it never had. Three undocumented client systems became read-only knowledge vaults in eight days. A patient-refund payout system reached me with 284 refunds stranded in two files: five days after first contact both files were resent, processed and paid, and the system now reports its own failures and rejects bad data at intake. It has since caught a failing file, and I told the client before they saw it.
I estimate accurately, which is what an ongoing client is really buying: 122.5 hours against a 130-hour budget, then 99.75 against 100 on the same client's next engagement. One automation I built was confirmed still running in production by a client vice president fourteen months after delivery.
I am good at the part before the work starts. Most of what I have shipped ran on a stack somebody else picked, several on tools I had not used before. Reading unfamiliar code, finding the answer where there is no documentation, and getting to a working change quickly is the skill underneath everything above. The AI in those systems is a component I chose for the job, never the product I was selling.
I work with one or two clients at a time, inside the business rather than at arm's length from it. I have been an asset to every company I have worked for, and what I build is meant to outlast me: vaults written for the next engineer, and a starter template written to be cloned.
I founded and ran YenkoDev, a registered Philippine software studio, until September 2026. On my own contracts I am the engineer who does the work.
Tell me what is breaking, or what nobody owns, and I will tell you how I would approach it and what I think it costs.