Connor McGowenCTO & Co-Founder · Kiro MoneyAgentic AI, built like infrastructure.
I’m the CTO and co-founder of Kiro Money, where we give banks and credit unions their own AI agent — inside their app, in their brand, grounded in their customers’ real account data.
I got here by climbing the stack: site reliability, then machine learning, then AI products.
- now
- CTO & Co-Founder, Kiro Money
- focus
- Agentic AI for regulated finance
- based
- San Francisco, CA
Ask about Connor, or pick a question below. Every answer runs through the same steps a production agent would: context, routing, a specialist, guardrails.
- Kiro Money is backed by
- U.S. National Science Foundation
- Founders Factory
- Aviva
- Recognized by
- Finovate
- American Bankers Association
- Poets&Quants
- Mayfield Divot AI List
01Now
AI is about to sit between banks and their customers. We make sure it’s the bank’s own.
People have started asking AI about their money — and connecting their accounts to do it. Every one of those conversations happens somewhere their bank can’t see.
Kiro Money gives each institution its own agent. It lives inside their app, speaks in their brand, works from real account data, and can put the institution’s own products to work — with the customer’s OK.
How the system thinks
- 01 · Context
Understands the customer
Reads the full financial picture — accounts, transactions, goals — in real time.
- 02 · Routing
Picks the right specialist
A multi-agent system hands each question to the agent built for it.
- 03 · Guardrails
Stays inside the lines
Holds to the institution’s compliance and topic boundaries. Always.
- 04 · Action
Closes the loop
Surfaces the right next step and lets the customer take it, right there.
- REST API
- for teams building their own experience
- White-label embed
- live inside digital banking, no core overhaul
- MCP connector
- the institution’s agent, inside the assistants customers already use
- young adults reached
- 6,000+
- saved or invested by community members
- $2M+
- from kickoff to live, in a published case study
- <2 wks
- user messages in that deployment’s first 60 days
- 8,000+
2026, so far
- Sep 2026Eko partners with Kiro to put an AI assistant alongside investing in digital banking
- Aug 2026Spinwheel partnership brings real-time debt data into the assistant
- Aug 2026Quiltt partnership adds a white-label assistant to connected bank data
- Jul 2026Heritage Hub Federal Credit Union launches an AI money coach with Kiro
- May 2026Live demo on stage at FinovateSpring 2026
02Trajectory
Seven years, one direction: up the stack.
Every layer I’ve worked on sits underneath the one I work on now. Reliability first, then models, then products — then the company.
2026 — Present
San Francisco, CA
Deciding what to build, and being accountable for it. Leading engineering for the embedded agentic intelligence layer for banks and credit unions. Backed by the U.S. National Science Foundation, Founders Factory and Aviva.
L3AI products
HIVE TA Technologies
AI Software Engineer
Nov 2023 — Apr 2026 · 2 yrs 6 mos
Remote
Turning models into software people actually use. Two and a half years building AI software at a startup.
L2Models
Levi Strauss & Co.
Machine Learning Engineer, QA
Mar 2022 — Apr 2023 · 1 yr 2 mos
San Francisco, CA
Getting models to hold up outside the notebook. Machine learning engineering inside a global enterprise.
L1Infrastructure
Blackhawk Network
Site Reliability Engineer
Jun 2019 — Mar 2022 · 2 yrs 10 mos
Pleasanton, CA
Reliability is a feature. Production is the only test that counts. Site reliability for a global payments company. Started as an intern while finishing my degree, then joined full-time.Intern Jun 2019 – Jun 2020, full-time from Jul 2020.
Education
San José State University
B.S., Network & System Administration · Dean’s Scholar
2015 – 2020
03Principles
How I build things that have to work.
- 01Principle
Reliability is the feature.
An agent that mostly works is a demo. In banking, the edge cases are the product — so I build for the failure modes first.
- 02Principle
Guardrails are architecture.
Scope, compliance and tone are designed into routing, tools and policy. They are not a filter bolted onto the output.
- 03Principle
Ground everything.
Answers come from the customer’s real accounts and the institution’s own knowledge. If it can’t be grounded, it shouldn’t be said.
- 04Principle
Finish in an action.
Advice that ends in a paragraph is content. The system should carry someone to the next step — and take it only with their OK.
04Toolbox
The stack, top to bottom.
What I reach for today, and the foundations underneath it that I still lean on every week. The tags under a row are specific things I’ve built with that layer.
AI systems
- 01Agents
- LangGraph
- LangChain
- Multi-agent orchestration
- Supervisor & sub-agents
- Parallel tool calls
- Durable execution
- Long-running agents
- Proactive agents
- Human-in-the-loop approvals
- 02Protocols
- MCP servers
- Remote MCP with OAuth
- Tool calling
- 03Models
- Gemini
- OpenAI
- Claude
- Qwen
- Gemma
- DeepSeek
- Open-weight models
- Model routing
- Fine-tuning (LoRA)
- Structured outputs
- 04Retrieval
- RAG
- Embeddings
- Vector search
- Web search
- Hybrid search & reranking
- Long-term memory
- Document understanding
- 05Reliability
- Guardrails
- Context engineering
- LLM tracing
- Eval suites in CI
- Prompt-injection defense
- Audit trails
Engineering underneath
- 06Backend
- Python
- FastAPI
- Pydantic
- SSE streaming
- 07Data
- MongoDB
- Pinecone
- 08Frontend
- TypeScript
- React
- Next.js
- 09Foundations
- Google Cloud
- Site reliability
- Observability
- Networking
05Contact
Building or backing agentic finance?
I’m always up for a conversation with investors, banks and credit unions, and engineers who care about making AI dependable.