Kevin Murphy is a product engineer focused on applied AI, based in Tempe, Arizona. He builds user-facing AI surfaces (streaming UI, agent loops, tool design, evals) on top of five years of production frontend at federal scale (Federal Reserve FedNow, IRS.gov, Michigan unemployment insurance) and three years on the ASU Mobile App on React Native and AWS. He holds a BA in cultural anthropology from the University of Vermont and is AWS Certified Developer Associate and AWS Solutions Architect Associate.

What does Kevin Murphy do?

Kevin builds AI-applied product surfaces end to end: the streaming UI, the agent loop, the tool design, and the evals. His shipped open-source artifacts include tablesalt (a CSV agent where the answer is the UI), streamfield (a React library for AI streams that do not look broken), grant-pilot (a multi-turn agent over federal grant data), fedbench (an LLM eval harness for grounded Q&A), fieldops-mcp (an MCP tool surface designed for agents), and mdx-corpus (a tiny npm package that turns an MDX directory into a retrieval-ready corpus).

Where has Kevin Murphy worked?

Kevin has worked at Deloitte (2021 to present, promoted to senior in 2023) leading frontend on federal-scale systems including the Federal Reserve FedNow onboarding portal, IRS.gov payment surfaces, and the State of Michigan unemployment insurance modernization. Before Deloitte he spent three years at Arizona State University as full-time staff on the ASU Mobile App using React Native and AWS, did two solo freelance SaaS builds during the pandemic, and started his career at the Smithsonian Institution updating the Folkways digital archive.

Kev-O is a grounded RAG chatbot trained on Kevin's own writing, deployed on his portfolio site at kev-o.kevinmurphywebdev.com and as inline punch-ins on individual blog posts and case studies. It uses Voyage embeddings + Claude streaming through the Vercel AI SDK, with hybrid BM25 + cross-encoder rerank retrieval over a corpus built from the MDX content on the same site. The infrastructure (corpus builder, retrieval pipeline, page-aware context injection) is documented in the "Building Kev-O" post.

What kind of roles is Kevin Murphy open to?

Kevin is open to Product Engineer or Applied AI roles at product-led teams in 2026. He is most interested in companies where the user-facing surface is the differentiator and where AI-applied work is shipped to real users, not just demos. He prefers contexts where one person can own a feature end to end (design, engineering, evals, observability) rather than narrow track roles.

How do I contact Kevin Murphy?

Email is the best path: kmurphywebdev@gmail.com. He is also on LinkedIn (midimurphdesigns), GitHub (midimurphdesigns), X (midimurph), and Bluesky (midimurph.bsky.social). His contact page at kevinmurphywebdev.com/contact has the canonical links.

about

Engineer with taste,
not artist who codes.

Currently open to Product Engineer / Applied AI roles at product-led teams

I'm Kevin Murphy, a product engineer based in Tempe. I build AI-applied product surfaces end-to-end: the streaming UI, the agent loop, the tool design, the evals. That work sits on top of five years shipping production frontend at federal scale (Federal Reserve FedNow, IRS.gov, Michigan UI), and three years building the ASU Mobile App on React Native and AWS as full-time ASU staff before that. I write the software for product-led teams who treat how the work feels to use as a competitive advantage, and I bring Claude Code daily plus MCP authoring experience to the engineering surface.

My origin is unusual for the role. I started in tech as a Smithsonian Institution intern in 2015 with a BA in cultural anthropology, updating the Folkways digital archive in the building that held the original vinyl pressing of This Land Is Your Land. From there the arc went through Arizona State University, two solo freelance SaaS builds during the pandemic, and Deloitte, where I came in mid-level in 2021 and was promoted to senior around 2023, leading frontend on federal-scale systems most people use without ever knowing the names.

That arc left me holding two halves of the work I rarely see in the same person: the analytical, systems-shaped engineering instinct, and the observational, interpretive instinct cultural anthropology trains for. I notice things on the human side of a build that purely technical training doesn't reach for, and I bring those notices back into the code. The site you're reading is one of the cleaner pieces of evidence.

AWS Certified Developer Associate and AWS Solutions Architect Associate. Both current.

what I bring

The case for hiring me, in five lines.

craft
Quality as a load-bearing skill, not a finishing touch.
Strict TypeScript, deliberate motion, accessibility built in, performance budgets defended, eval discipline on the AI side. The site you’re on is the proof artifact. Most engineers don’t ship the surface or the AI tooling this carefully because most engineers don’t treat either as part of the work.
scale
Production frontend at federal scale.
Since 2021, shipping federal-scale React systems at Deloitte that most people use without ever knowing the names: FedNow’s payment rail with the Federal Reserve, IRS.gov payment infrastructure, Michigan’s unemployment insurance modernization. Complex state, performance under load, design-system consistency, release trains under audit. The stakes were always real and the work always had to hold.
lens
A second instinct nobody else on the team brings.
A BA in cultural anthropology turns into applied anthropology on the job. Ethnographic patience, fieldwork instinct, attention to the gap between what users say and what they do. It’s the part of the role that doesn’t show up in a tech screen and consistently shows up in shipping quality.
tooling
Early on agentic AI, fluent on the daily driver.
Claude Code runs my whole loop. I’ve been building with agentic CLIs since the tooling was rough, and I have informed opinions about where the discipline is going. I bring fluency to that conversation, not curiosity.
depth
Senior reasoning across the whole stack.
TypeScript, React, Next.js, Node, Postgres, Supabase, Vercel. I can hold a system end-to-end and ship without a separate frontend team behind me. With AI tooling at 2026 levels, that combination ships at a quality bar small teams couldn’t reach five years ago, and that’s the configuration the work is moving toward.

what I'm looking for

The shape of the next role.

company
Product-led teams at AI-native scale-ups and design-forward dev-tools companies.
Two shapes I’d be excited to join, both equally. AI-native scale-ups in the Anthropic, Augment, Raspberry AI, Glean, Writer, Sierra, Decagon, Cresta shape, where the product is being built around AI from the ground up and a product engineer who can hold the agent stack, the eval loop, and the user-facing surface together has outsized leverage. And design-forward product / dev-tools companies in the Vercel, Linear, Resend, Clerk, Ashby, PostHog, WorkOS, Langfuse, Trigger.dev shape, where engineering quality and product taste are the competitive advantage. The right team, the right role, and the right project decide between them.
work
Product engineering on AI-native surfaces where the bar can't slip.
Roles where I own the surface customers actually touch. Streaming responses, agent loops that recover, tool designs that don't leak, components that feel intentional under traffic, paired with the orchestration and eval discipline behind them. Federal-scale audit pressure on Michigan UI, IRS.gov, and FedNow taught me what real-stakes engineering looks like, and I want that same bar applied to AI-applied product work, where most teams are still learning what production rigor means in this category. Applied AI engagements at the labs are a parallel track when the team and project are right.
moat
Companies positioned to win the AI era, not defend against it.
Real moats (proprietary data, regulated workflow lock-in, distribution that AI alone can't replicate) paired with monetization that scales with value delivered, not seats provisioned. Vertical SaaS in healthcare, legal, government services, and financial compliance is the canonical fit; single-feature SaaS without a moat is the canonical risk.
format
Remote, hybrid, or in-person. All three work.
I've spent years remote-first at Deloitte coordinating with offshore teammates in India and onshore colleagues across the country, so I know how to ship in distributed mode. I also know in-person collaboration on a strong team is its own kind of leverage and I'd show up for it. Pick the format that fits the role; I'll meet you there.
place
Tempe today; open to relocating for the right role.
I've built a life in Tempe (friends, gym, daily rhythm, cost of living that works) and remote from Tempe is the easy default. But I'd move for the right role. Bay Area, New York, anywhere a strong team is shipping work I'd be excited to do. Geography isn't the gate. The role, the team, and the project are.

commonly asked

Questions, answered straight.

Who is Kevin Murphy?: Kevin Murphy is a product engineer focused on applied AI, based in Tempe, Arizona. He builds user-facing AI surfaces (streaming UI, agent loops, tool design, evals) on top of five years of production frontend at federal scale (Federal Reserve FedNow, IRS.gov, Michigan unemployment insurance) and three years on the ASU Mobile App on React Native and AWS. He holds a BA in cultural anthropology from the University of Vermont and is AWS Certified Developer Associate and AWS Solutions Architect Associate.
What does Kevin Murphy do?: Kevin builds AI-applied product surfaces end to end: the streaming UI, the agent loop, the tool design, and the evals. His shipped open-source artifacts include tablesalt (a CSV agent where the answer is the UI), streamfield (a React library for AI streams that do not look broken), grant-pilot (a multi-turn agent over federal grant data), fedbench (an LLM eval harness for grounded Q&A), fieldops-mcp (an MCP tool surface designed for agents), and mdx-corpus (a tiny npm package that turns an MDX directory into a retrieval-ready corpus).
Where has Kevin Murphy worked?: Kevin has worked at Deloitte (2021 to present, promoted to senior in 2023) leading frontend on federal-scale systems including the Federal Reserve FedNow onboarding portal, IRS.gov payment surfaces, and the State of Michigan unemployment insurance modernization. Before Deloitte he spent three years at Arizona State University as full-time staff on the ASU Mobile App using React Native and AWS, did two solo freelance SaaS builds during the pandemic, and started his career at the Smithsonian Institution updating the Folkways digital archive.
What is Kev-O?: Kev-O is a grounded RAG chatbot trained on Kevin's own writing, deployed on his portfolio site at kev-o.kevinmurphywebdev.com and as inline punch-ins on individual blog posts and case studies. It uses Voyage embeddings + Claude streaming through the Vercel AI SDK, with hybrid BM25 + cross-encoder rerank retrieval over a corpus built from the MDX content on the same site. The infrastructure (corpus builder, retrieval pipeline, page-aware context injection) is documented in the "Building Kev-O" post.
What kind of roles is Kevin Murphy open to?: Kevin is open to Product Engineer or Applied AI roles at product-led teams in 2026. He is most interested in companies where the user-facing surface is the differentiator and where AI-applied work is shipped to real users, not just demos. He prefers contexts where one person can own a feature end to end (design, engineering, evals, observability) rather than narrow track roles.
How do I contact Kevin Murphy?: Email is the best path: kmurphywebdev@gmail.com. He is also on LinkedIn (midimurphdesigns), GitHub (midimurphdesigns), X (midimurph), and Bluesky (midimurph.bsky.social). His contact page at kevinmurphywebdev.com/contact has the canonical links.

off the clock

a guitar player — A '57 reissue Strat. A PRS CE24. A Taylor acoustic when nobody wants to be loud., a concert chaser — Spafford. Goose. Phish. If you know, you know., a blues believer — Jimi Hendrix is my hero in life. Albert King, Muddy Waters, and yes — John Mayer's blues chops are the real deal., a rock disciple — Led Zeppelin and Pink Floyd. The rest fills in around them., a jazz nerd — I'm there for the improvisation, not the names. Anyone who can actually solo over the changes., a funk evangelist — Cory Wong. The Meters. James Brown. Vincen García on the modern side., a jam-band convert — Long shows where the band actually goes somewhere. Setlist forensics on the drive home., a gym regular — Strength + zone-2 cardio. Boring works., a home cook — Stainless only. No grill. Top-shelf meal preps. Working on the butter-basted ribeye., a cat dad — His name is Ralph. He runs the house. I lease space., an animation enthusiast — Ralph Breaks the Internet. The Lorax. Comfort rewatches., a student of buddhism — Pema Chödrön. Thich Nhat Hanh. Eckhart Tolle's Power of Now on a re-read every couple years., a meditation practitioner — Daily-ish. Some days I just stare at the wall., a pickleball convert — Yes I bought the paddle. Yes it was worth it., a restaurant scout — Hidden Phoenix lists, road-trip detours, never the highway exits., a new-tech enthusiast — Claude Code is the daily driver. EVs and climate tech eat the rest of my attention., a technologist — Day-job senior engineer. Off-hours it's indie SaaS and whatever model dropped this week., a road-trip romantic — I-70 through Colorado. Day trip to the Grand Canyon., a lifelong tinkerer — Take it apart, put it back together, learn what was actually broken., a video-game lifer — Pokémon Blue on a Gameboy Color got me started. Now the Steam library's deep enough to be embarrassing., an RPG diehard — Started on Runescape in middle school. Now it's Elden Ring and BG3 eating weekends.