How I can help
Five kinds of work. Most engagements mix two of them. Start with what’s going on, and I’ll point you at the right one.
What’s going on?
- We have an agent pilot that’s stuck before production.Building agents
- Our app needs agents that people can see, stop and approve.Agentic and generative UI
- Our assistant answers confidently, and wrongly.RAG and agentic RAG
- We need senior judgement on agents a few hours a week.Fractional architecture
- Our users live in AG Grid and we want an agent in there with them.Agentic AG Grid
Building agents
I build agents that do real work inside your systems: they call your tools, ask before they act, and come with the tests that prove they still work after the next model change. Usually on b4.run and LangGraph, in your cloud.
- An agent in production, in your repo and your cloud.
- An eval suite that runs without an API key.
- Runbooks, tracing, and a handover to your team.
Usually four to eight weeksStack b4.run, LangGraph, AG-UI
Agentic and generative UI
The part people actually touch. Chat, threads, approvals, and interfaces the model assembles from your own components, in Angular or React, with threadplane, hashbrown, CopilotKit or assistant-ui.
- An interface where people can see, stop, approve and resume the agent.
- Generative UI built from your own design system.
- Streaming, durable threads that survive a refresh.
Usually four to ten weeksStack threadplane, hashbrown, CopilotKit, assistant-ui
RAG and agentic RAG
Answers grounded in your documents, with the source shown every time. When one search isn’t enough, an agent plans the retrieval, checks its own work, and iterates until the answer holds up.
- Ingestion and indexing over the content you already have.
- Citations on every answer, and permission-aware retrieval.
- An eval set that tells you when quality drifts.
Usually two weeks to prove it, then six to eightRecent emergency planning at IEM
Fractional architecture
A few hours a week of judgement for teams that don’t need a full-time hire: which agents to build, which protocols to bet on, how to evaluate them, and what to stop doing.
- Architecture reviews and design sessions with your team.
- A written point of view on agents, protocols and evals.
- Pairing on the hard parts, in your codebase.
Usually eight to thirty-two hours a monthFormat remote, every week
Agentic AG Grid
Agents that work inside the grid your users already live in: they filter, sort, group and edit through AG Grid’s own API, and every edit waits for approval. As a copilot beside the grid or as a grid inside the chat, on AG Grid Community or Enterprise.
- A copilot that drives your grid through its own API.
- Grids inside the chat, with edits that wait for approval.
- A performance baseline and an eval set for the grid.
Usually four to eight weeksStack AG Grid Community or Enterprise, threadplane
Side by side
If you’re comparing, this is the short version.
| Service | Typical length | You leave with | A good first step |
|---|---|---|---|
| Building agents | Four to eight weeks | An agent in production, its evals, runbooks | A 30-minute call about the one job the agent should do |
| Agentic and generative UI | Four to ten weeks | An interface people can watch, stop and correct | Send me a screenshot of where the agent lives today |
| RAG and agentic RAG | Two weeks to prove it, then six to eight | Grounded answers with citations, and an eval set | Twenty questions your users ask that it gets wrong |
| Fractional architecture | Eight to thirty-two hours a month | Decisions, reviews and a written point of view | One architecture review to see if we work well together |
| Agentic AG Grid | Four to eight weeks | An agent that drives your grid, with approval on every edit | A screenshot of your grid and the question your users ask it |
How working with me works
Short, fixed steps, so you always know what you’re paying for and what you’ll have at the end of it.
Week 0
A 30-minute call
You tell me what the agent is supposed to do. I tell you honestly whether I’m the right person.
Within a week
A written scope
What I’ll build, how we’ll know it works, what it costs. Fixed, in writing, before any work starts.
Two to six weeks
Build in sprints
Working software every week, in your repo, with the evals that prove it.
The last week
Handover
Your team owns it: the code, the docs, the runbook, and a walkthrough.
Buying from LiveLoveApp
I’m one person, so here is what your procurement and security teams will ask, answered up front.
A US company. LiveLoveApp, LLC, in Bend, Oregon. W-9 on request.
Your paper or mine. I have an MSA and an NDA, and I’m happy to sign yours.
Insured. Professional liability insurance, certificate on request.
Net-30. Invoices through your AP process.
Your environment. I work in your repo and your cloud. Your data stays where it is.
The IP is yours. Everything I write for you belongs to you.
Tell me what your agent is supposed to do.
Email brian@liveloveapp.com or book 30 minutes. I reply within a business day.