Systems that think.
Interfaces that don't.
I'm Azahid García — I build and evaluate AI systems: LLM pipelines, agentic workflows, and the evals that keep them honest. Founder of Strivark. Bias for clarity, performance, and craft.
I build things that think.
I'm an Applied AI Engineer with production experience shipping LLM-powered features, agentic workflows, and multi-agent orchestration systems in Python and TypeScript. I build AI products users actually rely on — not demos — with eval loops, structured output contracts, and observability that surfaces degradation before it reaches users.
Four years of enterprise consulting across LATAM and EMEA means I sit comfortably at the boundary between an ambiguous business problem and the engineering decision that resolves it. I work best in the gap between prototype and production — where rigor and craft both matter.
Currently building Strivark, a full-stack AI consulting platform with active enterprise clients. Based in Mexico City — working globally.
Play with what I build.
real interactive demos — no screenshots, no paragraphs. pick a tool and use it. this is the kind of thing I build for clients.
Selected work.
Four years inside enterprise consulting taught me how much billable time dies in repeatable operational work. Strivark routes that work to AI pipelines, freeing consultants for high-value reasoning.
visit ↗Single LLM calls fall apart on long-form, multi-step reasoning. This system chains specialized agents with retrieval and tool use, holding coherence across contexts that would break a single call.
private — write-up on requestAI agents systematically misuse tools — not because the code is broken, but because schema descriptions are ambiguous. Every systematic misfire mapped to description-level ambiguity; it reshaped how I write all tool contracts.
case study ↗Silent failures in LLM pipelines are invisible until they reach users. This system makes degradation loud: schemas fail explicitly, fallback frequency is instrumented, and every failed output logs its triggering input.
private — write-up on requestLet's build something worth building.
Open to applied AI and model evaluation roles, and hard problems at the frontier of AI. Based in Mexico City — ready to relocate. Reach out directly. I read every message personally.
azahid.garcia@gmail.comthe arc closes — the ánimas return and scatter upward