Questions before we talk.
Short answers about working together, pricing and how I keep LLM output reliable. Anything else fits in a 20-minute call.
Working together
Who will I work with?
With me, Carlos Salguero, directly. QANTARA is a one-person practice on purpose: no account managers and no handoffs. The person on the call is the person writing the code.
Where are you based, and which time zones do you cover?
Querétaro, Mexico (Central Time, UTC−6). I work with teams in Mexico and the US, in Spanish or English.
How do we start?
Book a 20-minute intro call or send a short brief. I reply within 48 hours with an honest read on fit and next steps.
Will you sign an NDA?
Yes. I'm happy to sign yours before we look at code or data.
Services and pricing
How much does each service cost?
Starting prices: LLM Reliability Review from US$3,000, Production AI Build from US$15,000, and AI WhatsApp Assistant from US$4,000 plus US$400/month. Every project gets a fixed price in writing before work starts.
Why do I see prices in MXN or USD?
Visitors in Mexico see prices in Mexican pesos; everyone else sees US dollars.
Should I start with a review or a build?
If you already have an LLM feature and aren't sure it can be trusted, start with a Reliability Review. If you know what you need and want it built for production, go straight to a build.
What happens after launch?
Ongoing support (evals, monitoring and fixes) is available from US$2,500/month. It's optional.
Technical
How do you stop the model from making things up?
Grounding first: the facts are extracted deterministically before any model call. Then evaluation against labeled cases, and a quality gate that decides whether each output is published, revised or escalated to a person.
Which models do you work with?
My production systems run on AWS Bedrock and Gemini. I choose the model per task and measure it, rather than committing to one vendor.
Do you also do backend and integration work?
Yes, as part of AI projects: Go, .NET, Spring Boot, Kafka and Salesforce/IICS integrations.
How do you handle our data and access?
With the least access the job needs, removed when the engagement ends. The security page has the details.
Accepting new projects
Have an LLM feature nobody fully trusts yet?
Twenty minutes is enough to tell whether a Reliability Review or a build makes sense.