AI Consulting & Engineering
Most AI projects fail in production, not in the demo. The prototype works. The board is excited. Then the API bill comes in at 10x the estimate, the model hallucinates in front of a customer, and the team spends a quarter debugging instead of shipping.
I help teams skip that phase. I bring 8+ years of production engineering, including three years as a Senior Engineer at Eventbrite, to build AI systems that are reliable enough to trust and cheap enough to run.
Three engagement models, depending on where you are:
Fractional AI Lead
You need AI leadership but not a full-time VP. I embed with your team 2 days per week to shape strategy, hire engineers, and architect production systems.
What you get after 90 days: An AI engineering org with clear standards, a defensible stack, a measurement layer, and a team that can ship without me.
- Hiring & team building. Define roles, screen candidates, and structure your AI engineering org
- Tech stack selection. Choose models, vendors, and infrastructure with rationale you can defend to your board
- Architecture reviews. Audit existing systems for reliability, cost, and latency before they hit production
- Eval pipeline setup. Build the measurement layer so you know whether your AI is actually working
Investment: Monthly retainer, scope-dependent based on team size, engagement depth, and hours per week. Contact for quote.
Zero-to-One Engineering
You have an idea and a deadline. I transform it into a working, testable MVP in weeks, not quarters.
What you get: A deployed, testable AI product with evaluation harness, CI/CD, and documentation your team can operate without me.
- Full-stack AI application. From prompt to production UI, deployed and accessible
- RAG / agentic workflows. Retrieval, tool use, and multi-step reasoning designed for your domain
- Deployment & CI/CD. Infrastructure that lets you ship updates without breaking what already works
- Evaluation harness from day one. You will never wonder "is it getting better or worse?"
See it in practice: I built celestino.ai from zero - a production voice agent with LiveKit, RAG grounding, tiered rate limiting, and graceful degradation - using this exact engagement model.
Production Engineering Sprint
Your AI works. It just costs too much. A focused 2-4 week engagement to review and optimize existing AI workloads for cost, latency, and quality.
What you get: A clear picture of where your AI spend is going, which optimizations are worth making, and a cost-reduced system with observability in place.
| What I have delivered | Result | |---|---| | Retrieval cost reduction at Eventbrite | 99.7% reduction, $15K/day to $40/month | | Vendor spend elimination | $60K/month saved | | Organic impression growth | +482% YoY |
- Eval pipelines & quality measurement. Define what "good" looks like and track it automatically
- LLM routing & vendor optimization. Send the right queries to the right models at the right price
- Cost reduction & latency improvement. Find the 80/20 wins that cut your bill without cutting quality
- Observability & monitoring. Debug "why did it say that?" in seconds, not hours
Why Me, Not a Consulting Firm
- I write the code. No slide decks that get handed to a junior developer. I architect and ship the system myself.
- I have done this at scale. Three years as a Senior Engineer at Eventbrite, building systems that served millions of users.
- I treat AI as engineering, not magic. Evals, cost controls, failure modes, and observability from day one. The same rigor I would apply to any production system.
- I build vendor off-ramps. No lock-in to a single model provider. The systems I build can switch models without rewriting application code.
Let's figure out the right engagement
Every engagement starts with a $300 strategy session: 60 minutes where we map your current state, define success, and scope the work. Fully credited toward your project if we proceed.
- 60-minute focused session
- Actionable next steps document
- Fully credited toward projects