Hey, I'm Carl.

I'm a Software Engineer with 11 years of experience. I help AI SaaS companies cut their LLM bills - including my own: I run Kuverly, where this approach removed 90% of our LLM spend without touching output quality. It's built on a Guard, Embed, Cascade system, three levers in order of impact.

How it works

1. Guard — stop paying for calls you don't need
A traditional ML classifier or circuit breaker filters requests before they reach an LLM. If a rule-based check can answer it, an LLM never gets asked.

2. Embed — replace LLM calls with something cheaper and reusable
Embeddings cost far less than a live LLM call and can be reused across requests. For routing, search, and inference tasks, they often eliminate the LLM call outright.

3. Cascade — pay flagship prices only for flagship problems
When an LLM call is unavoidable, route it to the cheapest model that clears your quality bar. Dropping one tier typically cuts that portion of your bill by 50%+; a full drop to the cheapest tier can cut it 80–90%. Redesigning the prompt itself to work on a lighter model compounds the savings further.

Stacked together, these three levers are how Kuverly's bill dropped 90% — not one trick, but removing calls first, replacing more with embeddings, and paying flagship rates only where they're actually earned.

Case Study

Case Study: Cutting LLM Costs by 90% Without Losing Quality

I run Kuverly, a social listening tool that ingests huge volumes of posts and surfaces only the relevant few to customers. That relevance judgment ran through a multi-shot LLM pipeline — several passes per post, tuned hard for quality. This is the process I used to cut our own LLM bill

Carl AndersonCarl Anderson

Testimonials

"He always considers the user experience and business objectives alongside the technical solution."

I worked closely with Carl for six months as his manager and was consistently impressed by his technical skills, initiative, and leadership. Carl took ownership of our frontend development efforts, providing strong technical guidance and helping drive important decisions. What sets Carl apart is his ability to think beyond the engineering problem itself. His insights were invaluable to the team, and I would highly recommend him to any organization seeking a talented engineer and technical leader.

— Fredrik Lindén

Find out what your LLM bill should actually cost

For LLM bills over $5000 a month, reach out to me to discuss your setup.

I can help you find:

• The true cost of your AI bills
• A ranked breakdown of opportunities to improve your high-cost LLM flows
• Places where unnecessary LLM calls are eating away at your margins