Skip to content
MN

About

Senior Applied AI Engineer

I lead agentic summarization, multi-turn evaluation, enterprise transcription, outbound voice AI, RAG, LLMOps, and responsible-AI work at Veterans United Home Loans. I’m also finishing a Master of Science in Computer Science at Georgia Tech, expected Spring 2027.

Career notes

Experience

  1. 01

    Senior Applied AI Engineer

    Feb 2024 — Present

    Veterans United Home Loans · St. Louis, Missouri

    • Architected a production AI note-taking platform for loan officers, preserving borrower context across team handoffs so a file keeps its history when it changes hands.
    • Served as technical lead for enterprise call-intelligence infrastructure built on Cisco CUBE, SIPREC, Amazon Chime, and AWS Call Analytics.
    • Led two engineers from architecture through production delivery of a Python and LiveKit outbound voice agent working previously unworked lead volume.
    • Built multi-turn evaluation suites with synthetic personas, expert labels, and human-calibrated LLM-as-judge scoring, catching regressions both before launch and after broad deployment.
    • Designed hybrid graph and semantic retrieval, established prompt and model lineage, and implemented the responsible-AI controls the AI portfolio runs on.
  2. 02

    Software Engineer

    May 2022 — Feb 2024

    Veterans United Home Loans · St. Louis, Missouri

    • Promoted domain-driven design across several projects, establishing consistent service and domain boundaries that made the codebase easier to reason about.
    • Designed systems that reduced Microsoft Dynamics CRM API reliance, contributing to faster lead creation and contact in the sales platform.
    • Resolved a lead-intake bottleneck that had been slowing processing, and standardized engineering and infrastructure processes across the team.
    • Developed NLP call-topic classifiers using scikit-learn, TensorFlow, and TF-IDF feature engineering — the evaluation habits that later carried into LLM work.
    • Initiated and led the company's first internal hackathon, and helped productionize a resulting solution that reduced third-party vendor costs.

Working toolkit

Technologies in context

Tools matter when they help the system hold together. These are grouped by the kinds of problems I’ve used them to solve rather than presented as a scoreboard.

Generative AI & agents

  • OpenAI
  • Anthropic
  • Amazon Bedrock
  • LangChain
  • LangGraph
  • LangSmith
  • Promptfoo
  • LiveKit
  • ElevenLabs
  • Claude Code
  • Tool calling
  • Structured outputs
  • Fine-tuning
  • Multi-turn evaluation

RAG & retrieval

  • Neo4j
  • Elasticsearch
  • FAISS
  • ChromaDB
  • Embeddings
  • Semantic search
  • Graph retrieval
  • Hybrid retrieval
  • Document chunking
  • Grounded generation

Cloud & LLMOps

  • AWS (Chime, SQS, Lambda, EC2, S3)
  • Kubernetes
  • Docker
  • MLflow
  • Snowflake
  • Bedrock Prompt Management
  • LangSmith tracing
  • LaunchDarkly

Software engineering

  • Python
  • C#
  • Java
  • FastAPI
  • SQL
  • Redis
  • Microsoft SQL Server
  • Distributed systems
  • Asynchronous processing
  • REST APIs
  • Microservices

ML, evaluation & governance

  • scikit-learn
  • TensorFlow
  • pandas
  • TF-IDF
  • F1 / precision / recall
  • LLM-as-judge
  • Okta RBAC
  • HashiCorp Vault
  • PII minimization
  • Llama Guard
  • Prompt-injection defenses

Domains

  • Finance & mortgage lending
  • Banking
  • Healthcare & medical technology
  • Legal technology

Learning

Education

Coursework →

  • Georgia Institute of Technology

    Expected Spring 2027

    M.S., Computer Science (OMSCS)

    Graduate study across machine learning, artificial intelligence, cognitive science, statistical evaluation, and software systems. Research includes evaluating RAG with Gemma, Qwen, and Llama models, comparing retrieval quality, task performance, and model-size tradeoffs.

  • University of Missouri

    May 2022

    B.S., Computer Science — Summa Cum Laude

    Columbia, Missouri.

Recognition

  • Summa Cum Laude, B.S. Computer Science, University of Missouri (2022).
  • Founder and lead of Veterans United Home Loans' first internal hackathon; a resulting solution was productionized and reduced third-party vendor costs.

Contact

Get in touch

The fastest way to reach me is email. If you’re building an AI team or trying to get an LLM system from prototype to production, I’d like to hear about it.

mikipux7@gmail.comMore ways to reach me