Professional Summary
Senior AI Engineer with 8+ years of software engineering experience building and operating production AI systems in healthcare and financial services. Designs agentic workflows, RAG and GraphRAG platforms on AWS Bedrock AgentCore and Azure AI Foundry, backed by Python/FastAPI services. Known for treating AI as an engineering discipline: measurable evaluation, layered guardrails, PHI protection (HIPAA, SOC 2), latency and token-cost control, and human approval for high-stakes actions.
Technical Skills
LLM & Prompt Engineering — Prompt design, structured outputs (JSON schema, Pydantic), tool/function calling, context management, prompt caching, model selection and routing by cost/quality
Agentic AI — LangGraph, CrewAI, AutoGen, Agent2Agent multi-agent orchestration, stateful workflows, human-in-the-loop approval, MCP, Salesforce Agentforce
RAG & Knowledge Systems — Ingestion pipelines, chunking, embeddings, hybrid search, reranking, citations, GraphRAG, knowledge graphs (Neo4j, Amazon Neptune), vector stores (Azure AI Search, Cosmos DB, pgvector)
Evaluation & Observability — Ragas, LangSmith, Azure AI Foundry evaluation, golden datasets, LLM-as-judge, regression gates in CI, tracing, token-cost and latency monitoring
Safety, Security & Governance — AWS Bedrock Guardrails, Azure AI Content Safety, prompt-injection defense, PHI de-identification, RBAC, audit logging, HIPAA, SOC 2
AWS — Bedrock, Bedrock AgentCore, SageMaker AI, Lambda, S3, IAM, CloudWatch
Azure — AI Foundry, OpenAI Service, AI Search, Cosmos DB, AI Content Safety, Machine Learning, Speech, Document Intelligence, Text Analytics
GCP & Other Platforms — Vertex AI, BigQuery, Cloud Composer, Dialogflow, Palantir Foundry
Voice AI — LiveKit, Deepgram (STT), ElevenLabs (TTS), Azure Speech, streaming responses, barge-in handling
Backend Engineering — Python, FastAPI, async programming, REST APIs, PostgreSQL, Redis, pytest, system design
Data Engineering — Databricks, PySpark, Snowflake, BigQuery, ETL/ELT pipelines, data quality and access control
DevOps & MLOps — Docker, Kubernetes, CI/CD (GitHub Actions), Terraform, model deployment and monitoring
Healthcare Domain — HIPAA, FHIR, Epic EHR integration, PHI handling, clinical workflow automation
Tools & Practices — Git, Jira, Agile/Scrum, Claude Code, Cursor, code review, technical documentation
Professional Experience
Senior AI Engineer, Hippocratic AI
01/2025 – 09/2026
  • •Architected and deployed clinical workflow agents on AWS Bedrock AgentCore and FastAPI, automating post-discharge follow-up and appointment outreach while replacing custom hosting, reducing deployment time from days to hours, and supporting 100K+ daily patient interactions.
  • •Built LangGraph workflows with typed tool calling and Epic FHIR APIs to read charts, schedule appointments, and document results; added human approval before consequential actions, preventing unapproved EHR changes while keeping clinician review under one minute.
  • •Built Python-based Model Context Protocol (MCP) servers that standardized secure, auditable access to Epic FHIR, scheduling, and knowledge-base tools across LangGraph, AutoGen, and CrewAI agents, cutting new integration onboarding from weeks to days.
  • •Built an Agent2Agent clinical workflow orchestration solution with AutoGen and CrewAI, coordinating intake, scheduling, benefits, and escalation agents; in internal testing on representative workflows, the multi-agent design increased task completion from 70% with a single overloaded prompt to 90%.
  • •Built and deployed production RAG pipelines with Databricks and PySpark to process clinical protocols and policies, integrating hybrid search, reranking, and source citations to reduce hallucinations by 20% versus baseline in internal evaluations
  • •Built and evaluated a GraphRAG system that modeled clinical conditions, medications, and care pathways in Neo4j, assessed Amazon Neptune, and combined graph traversal with vector search to improve accuracy on multi-hop questions by 25% versus vector-only retrieval.
  • •Reduced inference costs by 40% versus an all-large-model baseline while maintaining quality targets by benchmarking Bedrock foundation models, training SageMaker AI intent and triage classifiers, and routing tasks by complexity.
  • •Created an LLM evaluation framework with Ragas (faithfulness, context precision, answer relevancy) and LangSmith tracing, run as a release gate in CI, plus token-cost dashboards per workflow. Raised faithfulness from 82% to 94% and caught prompt regressions before production
  • •Implemented AWS Bedrock Guardrails with PHI redaction, denied-topic controls, and grounding checks, alongside patient-level authorization, audit logging, and human escalation for low-confidence responses, supporting customer security and compliance reviews while maintaining zero reported PHI incidents during the project period.
  • •Developed and launched a Salesforce Agentforce patient engagement agent integrated with clinical APIs, automating personalized outreach, increasing patient response rates 25% from the pre-launch baseline, and reducing manual follow-up.
  • AI Engineer, Zus Health
    07/2023 – 12/2024
  • •Built a governed Azure AI Foundry workflow to manage model deployments, prompt iterations, and evaluation runs, exposed AI features through FastAPI services, and reduced average release cycle time by 30% compared with the prior process.
  • •Built a secure RAG system on Azure AI Search for patient records and FHIR documents, reducing irrelevant-context answers and supporting 50K+ daily queries at under 2-second p95 latency
  • •Cosmos DB vector search: Stored embeddings and conversation memory in Azure Cosmos DB vector search for low-latency lookups alongside operational data, avoiding a separate vector database to run and secure.
  • •Azure OpenAI: Integrated Azure OpenAI Service deployments for summarization and clinical Q&A with structured outputs; handled rate limits with retries, request queuing, and response caching to keep quality stable under peak load.
  • •Reduced LLM spend by 30% versus using a single provider by integrating Amazon Bedrock behind a common abstraction layer, enabling outage fallback and routing suitable workloads to lower-cost models.
  • •Applied Azure AI Content Safety (including prompt-injection shields), input/output filtering, and PHI de-identification before logging. Passed HIPAA and SOC 2 reviews with no PHI incidents
  • •Built a real-time voice assistant using LiveKit, Deepgram, ElevenLabs, and Azure Speech fallback, streaming LLM responses directly to text-to-speech with interruption handling to support natural, responsive conversations and achieve sub-800 ms time-to-first-audio under internal testing conditions.
  • •Built golden datasets and Foundry evaluation runs to measure retrieval and answer quality, then trained and deployed a lightweight reranker through Azure Machine Learning, improving retrieval accuracy by 18% over the baseline on the evaluation dataset.
  • •Built Snowflake pipelines that normalized FHIR data for reliable retrieval and evaluation, while enforcing masking policies and role-based access controls to secure sensitive fields and enable compliant data use.
  • AI Software Engineer, Codat
    01/2021 – 06/2023
  • •Built and integrated AI, document-processing, conversational, and data-engineering solutions using Azure Cognitive Services, Document Intelligence, Vertex AI, BigQuery, Cloud Composer, Palantir Foundry, Bot Framework, and Dialogflow, automating workflows and improving data accessibility and user support.
  • •Built and deployed an AI-powered invoice and statement processing pipeline with Azure Document Intelligence, Text Analytics, Vertex AI, and RAG, processing 200K+ documents monthly, achieving 95%+ extraction accuracy on a labeled evaluation set and reducing manual review effort by 50%.
  • •Delivered customer-facing assistants using Microsoft Bot Framework and Google Dialogflow to answer finance workflow questions, serving 20K+ monthly users and deflecting 30% of support requests.
  • •Orchestrated pipelines with BigQuery, Cloud Composer, and Palantir Foundry, adding data quality checks, access controls, and audit trails to meet financial compliance requirements.
  • Software Developer, Google
    01/2018 – 12/2020
  • •Developed, containerized, and deployed Python applications using Docker, Kubernetes, CI/CD pipelines, and SQL.
  • •Backend APIs: Built and maintained Python APIs handling millions of requests per day at 99.9% availability; tuned database queries and data structures to cut latency by 30%.
  • •Containerized services with Docker and Kubernetes and improved CI/CD pipelines, shortening release time by 40%
  • Selected Projects
    MedLens Clinical Knowledge Assistant
  • •RAG assistant that finds answers across patient records and clinical guidelines with citations, patient-level authorization, and human review for uncertain results.
  • •Tech: Python, FastAPI, embeddings, reranking, Bedrock Guardrails
  • CareBridge Orchestrator
  • •Agent that gathers information across multiple systems and asks for explicit approval before consequential actions.
  • •Tech: Bedrock AgentCore, LangGraph, FHIR
  • Northwind Care Navigator
  • •Benefits and policy copilot with safety filters and quality monitoring.
  • •Tech: Azure AI Foundry, Azure OpenAI, Azure AI Search
  • Education
    Bachelor of Science, Computer Science, Texas Tech University
    08/2013 – 12/2017