Matthew Aragaw

About Matthew Aragaw

Software engineer and founder focused on building applied AI systems, LLM infrastructure, and distributed backend platforms.My work centers on designing AI agents, agentic workflows, and multi-agent systems that allow large language models to interact safely with real-world software systems. I’m particularly interested in architectures that combine LLM reasoning with deterministic execution layers, enabling reliable tool use, tool calling, and workflow orchestration across complex software environments. Extensive experience building distributed systems, scalable backend architectures, and developer platforms that power production AI systems. My work often involves designing platform APIs, SDKs, and developer tooling that make advanced infrastructure intuitive.Recently I’ve been focused on LLM infrastructure and agent orchestration, including building evaluation pipelines, agent harnesses, and feedback loops that allow AI systems to improve through human-in-the-loop evaluation and reinforcement learning workflows.I enjoy designing systems that operate reliably at scale, with strong focus on observability, reliability engineering, and production debugging. Much of my work involves building internal platforms and automation systems that ship AI-powered products quickly and safely.Languages: Python • C++ • Go • Java • TypeScriptAreas:Applied AI • LLM systems • AI agents • Agent orchestration • Prompt engineering • LLM evaluation • AI infrastructure • Distributed systems • Scalable backend architecture • Workflow orchestration • API design • Developer platforms • Platform engineering • Observability • Reliability engineering • Cloud infrastructure • Kubernetes

Experience

1. Founder

   Stealth

   Sep 2024 — Present

   Architected distributed microservice platform deployed over AWS ECS Fargate service mesh, powering strategy orchestration and cross-chain automated execution, supported by async SNS+SQS worker architecture and persisted via PostgreSQL/Redis. Implemented in Python FastAPI- Architected event-driven execution engine with unified policy resolution, concurrency-safe capital allocation, and SQS-backed parallel trade routing across multiple execution venues- Architected an LLM-driven workflow generation platform backed by Bedrock+S3 RAG pipeline, enabling creation of automated execution workflows through conversational chat UI- Designed a constrained SDK-based tool-calling framework to allow agents to safely interact with underlying execution infrastructure- Built a multi-layer validation pipeline including sandbox execution, backtesting simulation across historical data, and walk-forward validation for overfitting detection- Designed ERC-4337 backed smart account infrastructure and session key validation managers, enabling non-custodial automated asset management enforced on-chain.

Education

• UC Santa Barbara

  Bachelor of Science - BS, Computer Engineering

  2018 — 2022