Skip to main content

Sergio “Chekos” Sánchez · working field notebook

Complicated work should be easier to see, trust, and act on.

I build public-serving data systems, agent tools that keep people in control, hardware and software for small businesses, and technical education in Spanish.

Index A

Four practices, one through-line

Make useful systems legible enough for people to trust and change.

  1. 01

    Public-serving data & decision systems

    Infrastructure, analytics, and AI operations that make consequential work easier to inspect and act on.

  2. 02

    Agent-native tools & operating systems

    Tools that hold context, expose evidence, and leave judgment with the person using them.

  3. 03

    Product, hardware & small-business software

    Physical and digital products designed around real operating constraints, not imaginary scale.

  4. 04

    Bilingual teaching, writing & community

    Spanish-first technical education that makes practical knowledge useful without flattening its complexity.

Field records 01–04

Proof of depth, across the practice

TalkingPoints is the deepest record here, not the whole story. These four systems show the recurring shape of the work: hard constraints, useful artifacts, visible proof.

01Deepest record2022–2026

TalkingPoints

Public systems

Problem
A growing multilingual education platform needed data systems it could trust and afford.
My role
Joined the early data team and grew into a senior role across infrastructure, analytics, APIs, research, and AI operations.
Shipped
Six connected systems, from the warehouse foundation to partner analytics and evaluation loops.
Proof
70% lower Snowflake cost; self-service analytics; a new attendance product data surface.
Open field record →
−70%Snowflake cost
6connected case studies
Recorded outcomes · TalkingPoints
02Operating system2025–present

Nahual + Garita

Agent systems

Problem
Work spread across many repositories and agents kept losing its shared context.
My role
Designed and built the project-memory, capability, and GitHub-signal layers.
Shipped
A reviewed registry and receipt system connected to a signal router for public repositories.
Proof
One inspectable operating picture across local agents, web, Telegram, and GitHub.
Open field record →
repositoriesGarita signalsNahual reviewagents + Sergio
Evidence moves; judgment stays explicit.
03Hardware in the loop2024–present

Relod

Product + hardware

Problem
Food businesses need a quiet way to know what will run out before it does.
My role
Co-founder; forecasting, application, and data-system layers.
Shipped
Sensor-equipped containers, a live operations view, and reorder forecasting.
Proof
A working hardware system moved from prototype into café and bar pilots.
Open field record →
Relod sensor lid mounted on a clear coffee-bean container during a working pilot
Working sensor container · pilot hardware
04Spanish-first by design2018–present

Tacos de Datos

Teaching + community

Problem
Practical data and AI knowledge is still too often gated by English and insider language.
My role
Founder, writer, teacher, and community builder.
Shipped
Essays, tutorials, workshops, courses, and a welcoming technical community.
Proof
Publishing and teaching useful technical ideas in Spanish since 2018.
Open field record →
“No necesitas aprender inglés para que una idea técnica te resulte útil.”
Teaching premise · Tacos de Datos

Browse the complete practice-led portfolio

Live leaf · reviewed Jul 10, 2026

From the workbench

A small public-safe projection from Nahual—not the private project ledger. Every item is allowlisted, source-linked, and frozen into the static site after review.

soyserg.io

The field-notebook revamp connects a practice-led portfolio, reviewed workbench notes, and route-level impact estimates.

Public surfacePublic site

Snapshot · Reviewed public fields from the Nahual project registry, verified against linked public sites and GitHub sources.

Marginalia · three principles

Beliefs with receipts

Fewer slogans. Each principle points to work where it had consequences.

Technical ink study of a bird
P3

Build for real constraints.

Relod learned from working containers and pilot kitchens, not imagined enterprise scale.

See the Relod field record

Good reasons to reach out

Start with the interesting, difficult part.

The interesting problems usually sit where public purpose, human judgment, and real-world constraints collide. If that is what is in front of you, the contact page has a few concrete places to start.

Start a conversation →