David Beal

About David Beal

I build AI systems that work in production, not just in demos — specifically, custom multi-agent architectures for scientific-grade medical decision support.My focus is designing highly agentic systems where accuracy is non-negotiable: architectures that decompose complex clinical reasoning across specialized AI agents, evaluation pipelines that catch hallucinations before they reach a patient, and knowledge-graph-backed systems where every AI output is grounded in structured, verifiable medical data.What I do day to day- Design and ship custom multi-agent orchestration systems with tool-calling, state persistence, and intelligent routing across 15+ specialized medical AI agents- Build document intelligence pipelines that extract structured biomarker data from lab reports, genetic panels, and clinical documents with measurable, auditable accuracy- Architect graph-based medical knowledge systems (Neo4j) that give AI agents deep patient context — observations, conditions, medications, genetics — normalized to LOINC and RxNorm standards- Implement rigorous AI output evaluation: extraction accuracy scoring, variance testing, regression detection, and ground-truth validation pipelines- Debug the hardest problems in medical AI: prompt drift, extraction edge cases, agent routing failures, and the long tail of LLM misbehavior where clinical safety is at stake- Operate AI infrastructure at scale: Kubernetes on GKE, vector search (Qdrant), FastAPI backends, Next.js frontendsBefore this, I spent 12+ years as a systems engineer at Bloomberg, Google, and Bank of America, building the kind of low-level, high-reliability software (C/C++, distributed systems, security) that makes me deeply skeptical of AI output by default — which turns out to be the most important skill when building AI systems that give medical advice.Python, TypeScript, Neo4j, PostgreSQL, Redis, Qdrant, Kubernetes, GCP, FastAPI, Next.js, LangFuse.

Experience

1. Sr Director of Ai

   Wethosai

   Sep 2024 — Present · Utah County, UT, US

   Leading AI architecture and R&D for collaborative intelligence platform serving enterprise customers• Designing multi-agent systems for enterprise knowledge workflows using LLM orchestration and tool-calling patterns• Building evaluation frameworks to validate AI output quality across diverse enterprise use cases• Kubernetes/GCP infrastructure for scalable AI workloads

Education

• Pasadena City College

  Physics

• Los Angeles City College

  Mathematics

Skills

• Iphone Application Development
• Oracle
• Firmware
• Windows
• Bash
• Gnu Make
• Swig
• Web Applications
• Databases
• Network Programming
• Algorithms
• Php
• Computer Vision
• Rapid Prototyping
• C++
• Http
• Objective-C
• Bloomberg Terminal
• Embedded Systems
• C
• Optimization
• Sqlite
• Neural Networks
• Postgresql
• Genetic Algorithms
• Flex
• Unit Testing
• Python
• Django
• Unix Shell Scripting
• Java
• C#
• Multithreading
• Machine Learning
• Assembly
• Computer Graphics
• Vi
• Opengl
• Unix
• User Experience