Software engineer · Austin, TX

I build software
with AI.

I’ve been breaking and fixing systems since 2016, from IT support and cybersecurity to data and software.

Experience at
Tesla Ford Trexoros
AI-generated. The judge is fictional. A promo video with no camera Read the case study →

Nine calls became two

An AI agent needed nine tool calls to draft one small campaign. I rebuilt its tools and ran 595 controlled runs to test the change.

Read the case study →
agent · create campaign
User: “Create a campaign with an article, a LinkedIn post and an Instagram post. Submit everything for review.”
    MCP calls0
    Model turns0
    Tokens sent0

    Real traces from my benchmark: same model, same task.

    73%less median context per task
    (DeepSeek V4.1 Flash)
    98.6%Haiku task success,
    up from 87.5%
    0Off-limits actions
    in sandboxed runs

    Some of my work

    All 19 projects →

    How I can help

    01

    Build software.

    Web apps, internal tools, APIs and integrations.

    • Web apps and internal tools
    • APIs and integrations
    • AI features inside real products
    • MCP servers and agent tools
    02

    Automate work.

    Workflows and scripts that remove repeat tasks. Some of them use AI.

    • Scripts and scheduled jobs
    • Connecting the tools you already use
    • AI steps with a person approving the result
    • Tests that catch breakage early
    03

    Audit and fix.

    I review a system you already have. I find what fails and what costs too much. You get a ranked list of fixes.

    • Code and workflow review
    • Failures sorted by cause
    • Cost and speed breakdown
    • Agent and MCP trace review
    04

    Technology research.

    I compare tools, models or vendors. You get a written recommendation with evidence.

    • Tool and vendor comparisons
    • Model tests on your own tasks
    • Before and after measurements
    • A written list of limits

    My toolkit

    Build

    • TypeScriptApps and agents
    • PythonML, data, APIs
    • ReactInterfaces
    • Next.jsFull-stack apps
    • FastAPIServices

    AI & agents

    • MCPTool contracts
    • Claude CodeAgent host
    • EvalsBenchmarks
    • Vision modelsCreative pipelines
    • OpenTelemetryLLM tracing

    Data & ML

    • PostgreSQLProduct data
    • DuckDBAnalytics
    • AirflowData pipelines
    • PyTorchDeep learning
    • Boosted treesCatBoost, XGBoost

    Ship

    • DockerEnvironments
    • CloudflareWorkers, D1
    • PlaywrightBrowser tests
    • VitestContract tests

    Experience

    My story →
    Trexoros

    Technical Lead

    Building software and AI workflows: product, backend, agent tools, integrations and testing.

    Ford

    Product Manager Intern, Software Engineering

    Internal AI tools, reusable workflows, and a 130 person hackathon I organized.

    Tesla

    Data Engineer Intern

    Data work on the Cybertruck and Model Y: pipelines and quality tools, and helping engineers use AI in practical ways.

    Arizona State University

    B.S. Computer Science

    Computer science fundamentals behind the practical work.

    Earlier: cybersecurity at Fannie Mae, data science at PFES, software engineering at ProPharma, robotics teaching at Codex, and IT at Planet Pharma. See all

    “He ramped quickly, worked independently, and knew when to pull in help.”
    Internship feedback, 2026. Anonymized excerpt.

    Let’s talk

    Tell me about the workflow or the idea. I read every message myself.

    stombya@gmail.com