Dallas / Building with agents

I build software
that can act.

I’m Adrian—a lead backend and cloud engineer turning agents, APIs, and infrastructure into useful systems with clear boundaries.

See how I build↓

Pixel-art developer directing software agents from a neon Dallas studio — DIRECT / DELEGATE / VERIFY / SHIP

PORTFOLIO
’26

01 / AI & automated products

Useful software where AI has a defined job, evidence, and a path to production.

Systems
online.

Pixel-art agent team verifying evidence and assembling a resume on a CRT

001 Live site

AI WORKFLOW / PRIVACY-FIRST CAREER TOOL / 2026

ResumePilot

An evidence-first resume workspace that turns verified career facts into tailored outputs while preserving provenance, ownership, and human review.

Amazon Bedrock
React
Python
AWS
Terraform

Open ResumePilot ↗ View source ↗

resumepilot.soadrian.com

Pixel-art monitoring agents watching market signals from a retro command room

002 Live site

AUTOMATED MONITORING / RISK GUARDRAILS / FULL STACK

smart-margin

A continuously running drawdown monitor that observes market signals, evaluates explicit risk policy, and alerts people before thresholds become surprises.

Visit site ↗ View source ↗

smartmargin.soadrian.com

> shaping complex systems
  into clear interfaces

01 give the agent a job
02 constrain the action
03 verify the result

STATUS: HUMAN IN CONTROL_

NEXT SYSTEM / ALWAYS LEARNING

The next useful agent

I’m interested in agents that do real work inside well-designed systems—not demos that stop at the prompt box.

Start a conversation ↗

02 / Agentic development

Agents need
an operating model.

I use AI as an engineering collaborator: give it bounded context, connect it to capable tools, make its work observable, and keep consequential decisions reviewable.

DIRECT

Start with intent

Translate an ambiguous goal into a concrete task, the context needed to solve it, and an explicit definition of done.

DELEGATE

Give tools, not magic

Let agents inspect code, call APIs, and operate cloud workflows inside permissions designed for the job.

VERIFY

Make proof part of the loop

Use tests, source evidence, policy checks, and human review so speed never depends on pretending the model is infallible.

Current deliveryAWS Kiro + Amazon Bedrock

Engineering leverageAmazon Q + GitHub Copilot

Agent workflowsCrewAI + Claude

Human feedbackSageMaker + A2I

03 / Engineering judgment

The agent is fast.
The judgment is human.

I bring years of backend, API, and cloud delivery to agentic development. The model can accelerate the work; architecture, security, accountability, and the final call still need an engineer.

Production depth

Scalable APIs, event-driven systems, and cloud platforms across healthcare, fintech, and enterprise modernization.

Guardrails by design

Permissions, privacy, failure modes, and observability belong in the product—not in a cleanup pass.

Measured leverage

Use agents where they shorten feedback loops, then prove the result with tests, evidence, and review.

04 / Get in touch

Have work worth
delegating well?

adrian@rydeas.com↗ GitHub ↗

I build softwarethat can act.

Systemsonline.