Skip to content
Santiago GerickePortfolio  ·  2026

↳ Inference complete  ·  0.99 confidence

AI engineer
who ships.

Founded a brand at 18. Director of AI at TechTorch — co-built the practice 1→20, shipping agents into Fortune 500 production.

Based in SpainOpen to Bay Area
↓scroll01 / 05

By the numbers

Selected work · 2026

Three that
matter.

Ten more that shipped. What follows is the cut that carries the most weight — enterprise AI in production, an open protocol, and an autonomous workforce for DTC.

  1. 01TechTorch90%+
  2. 02AINM12
  3. 03Tayliq12+

Chapter 01TechTorch

TechTorch AI Quoting Agent

Enterprise sales teams spend hours building quotes in legacy Salesforce CPQ — and get them wrong almost as often as they get them right.

  • Multi-agent architecture with dedicated roles per quote stage
  • Novel Apex-native integration — CPQ operations run as native Salesforce transactions, not brittle service-router calls
  • Co-architected end-to-end: model selection, tool schemas, state machine, deployment
  • Live at a $2B cybersecurity company + in testing at 10 more enterprise clients
  • Anthropic Claude
  • OpenAI
  • Multi-Agent Architecture
  • Salesforce CPQ
  • Apex
  • Tool Calling
  • RAG
  • TypeScript
techtorch / quoting-agent.log
> agent.route("quote.assemble")
  └─ LLM: plan(actions)            ✓ 312ms
  └─ tool: cpq.query(product)      ✓ 68ms
  └─ tool: cpq.apex.addLine(...)   ✓ 41ms   (native tx)
  └─ tool: cpq.apex.applyDiscount  ✓ 38ms
  └─ LLM: verify(quote)            ✓ 287ms
  └─ tool: cpq.submit              ✓ 52ms
< DONE · 798ms · 1 tx · 0 retries
  • 90%+Faster quote creation
  • 80%+Faster reporting
  • $6–8MRun-rate impact
  • 20AI team 1 →

Chapter 02AINM

AINM — AI-Native Messaging Protocol

MCP covers tools. A2A covers agents talking to agents. AG-UI covers humans talking to agents. Nothing covers a thread where all three share the room.

  • Federated, AI-native messaging — humans and agents are equal participants in every thread
  • 12 formal specifications covering identity, transport, threading, memory, and compatibility
  • AsyncAPI 3.1 · JSON Schema 2020-12 · Ed25519 signing · mutual TLS · RFC 8785
  • Reference implementations: Rust (Axum) server · iOS (SwiftUI) app · Node.js SDK
  • Rust
  • Axum
  • Swift
  • SwiftUI
  • Node.js
  • TypeScript
  • AsyncAPI 3.1
  • JSON Schema
Visit
ainm / spec-index.md
  • 01 · identity
  • 02 · transport
  • 03 · threading
  • 04 · streaming
  • 05 · memory
  • 06 · interactive content
  • 07 · agent cards
  • 08 · tool visibility
  • 09 · security
  • 10 · signing (Ed25519)
  • 11 · compatibility
  • 12 · federation
  • 12Formal specs
  • 3Reference impls
  • MITLicense
  • Pre-v1.0Status

Chapter 03Tayliq

Tayliq

DTC brands drown in support tickets. Existing tools triage — Tayliq acts. It watches shipments, drafts in brand voice, processes refunds, and only pings you when a human is truly needed.

  • Agent #1 (Customer Support) is live today — Agents #2 and #3 (Ops, Finance) are next
  • Watches every parcel across DHL, GLS, Correos, UPS, FedEx, and SEUR
  • Drafts in 12+ languages in brand voice, auto-refunds, suggests size swaps before money leaves the account
  • Origin: the AI support agent I built for Shameless Collective — real traffic on a ~$1M DTC brand — productized for any Shopify merchant
  • GPT-4
  • Anthropic Claude
  • Shopify GraphQL API
  • Remix
  • App Bridge
  • App Proxy
  • Polaris
  • ElevenLabs
Visit
tayliq / workforce.manifest
  • 01Customer SupportLIVE
  • 02OperationsNEXT
  • 03FinanceNEXT
  • 12+Languages
  • 6Carriers watched
  • ShopifyBuilt inside
  • SOC 2 · GDPRCompliance

End notes · 09 projects

Also shipped.

Earlier systems — multi-agent, classification, voice, end-to-end automation. Each one sharpened a tool I later brought to the tentpoles.

  1. 01

    Multi-Agent AWS Bedrock System

    Designed a multi-agent AI system for intelligent recommendations on AWS

    • AWS Bedrock
    • AWS Lambda
    • Amazon S3
  2. 02

    Gmail AI Response System

    Created an AI system that monitors, categorizes, and responds to emails autonomously

    • Python
    • OpenAI API
    • Gmail API
  3. 03

    Multi-Class Text Classification Model

    Fine-tuned a DistilBERT model for multi-category classification on SageMaker

    • AWS Lambda
    • AWS API Gateway
    • Hugging Face
  4. 04

    Duolingo Clone

    Built a language learning platform with AI-generated voices

    • React
    • Next.js
    • DrizzleORM
  5. 05

    Shameless Returns

    Automated the returns workflow with real-time carrier tracking

    • Next.js
    • React
    • Tailwind CSS
  6. 06

    Shameless Locator

    Built a real-time order tracker with carrier API integration

    • Next.js
    • Tailwind CSS
    • Vercel
  7. 07

    Shameless Website - Shopify Clone

    Built a full e-commerce platform with auth, payments, and serverless data

    • Next.js
    • Tailwind CSS
    • Typescript
  8. 08

    Order Fulfillment Automation

    Automated order fulfillment from Shopify to carrier systems

    • Python
    • Shopify API
    • SOAP API
  9. 09

    Monthly Invoice Automation

    Automated monthly invoice generation and email dispatch

    • Python
    • Shopify API
    • ReportLab

The path · 08 milestones

How I got here.

  1. 2016 – Present

    Founded at 18

    Co-founded Shameless Collective while entering university. Built the whole stack myself — Shopify storefront, ops, finance, and the first AI customer support agent. On pace for $1M in 2026 across 45+ markets.

  2. 2016 – 2021

    Engineer + MBA

    Industrial Engineering at ICAI (double merit scholarships from the Comunidad de Madrid) and an MBA at ICADE. Found software along the way — never looked back.

  3. 2021

    Strategy at Kearney

    Joined a global top-tier management consultancy. Learned how Fortune 500s actually make decisions before I started building AI for them.

  4. 2022 – 2023

    Data Science at Metyis

    Senior Data Scientist. Architected the pricing optimization algorithm for Spain's largest independent brewer — driving pricing strategy across the portfolio.

  5. 2023 – 2025

    Shipping AI at Monstarlab

    Senior Data & AI Associate. Led enterprise AI delivery — multi-agent systems, classification, RAG, voice. Full spectrum, real clients.

  6. 2025

    Go Exponential Fellow

    Selected into a Spain → Silicon Valley fellowship (<2% acceptance), backed by Fundación Barrié and the Spain-US Chamber of Commerce.

  7. 2025 – Present

    Director of AI at TechTorch

    Recruited to build TechTorch's AI practice. Co-built the team from 1 to 20 in 7 months. Architected the flagship AI Quoting Agent — live at a $2B cybersecurity company and in testing at 10+ enterprise clients. Run-rate impact: $6–8M annually.

  8. 2026

    Open-source infrastructure

    Published AINM — an open-source AI-native messaging protocol. 12 specs, Rust + iOS + Node.js reference implementations. Sits alongside MCP and A2A in the emerging agent stack.

Get in touch · 04 channels

Let's build.

Santiago Gericke — AI engineer who ships.

Confidence: 0.99 · based in Spain · open to Bay Area.

By the numbers

  • 10+ years building software
  • 15+ production AI systems shipped
  • $1M brand co-founded at 18 (Shameless Collective)

Selected work

TechTorch AI Quoting Agent

Flagship enterprise AI agent running in Fortune 500 production

Flagship AI agent at TechTorch — live in production at a $2B cybersecurity company (Fortune 500) and in testing at 10+ additional enterprise clients. I co-architected it end-to-end: model selection, tool schemas, state management, and a novel Apex-native integration into Salesforce CPQ — running CPQ operations as native Salesforce transactions rather than the industry-standard service-router pattern. Measurable impact across deployments: 90%+ faster quote creation, 80%+ faster reporting, and $6–8M annual run-rate savings.

  • Anthropic Claude
  • OpenAI
  • Multi-Agent Architecture
  • Salesforce CPQ
  • Apex
  • Tool Calling
  • RAG
  • TypeScript
  • Python
  • PostgreSQL
View Project

AINM — AI-Native Messaging Protocol

Open-source protocol for human + agent + tool messaging

An open, federated messaging protocol designed from the ground up for AI agent communication — where humans and agents are equal participants in a thread. Scope: 12 formal specification documents covering identity, security, transport, threading, streaming, memory, interactive content, agent cards, tool visibility, and compatibility layers. Built on AsyncAPI 3.1, JSON Schema 2020-12, Ed25519 signing, mutual TLS, and RFC 8785 canonical JSON. Reference implementations: Rust (Axum) server, iOS (SwiftUI) app, and Node.js SDK. Fills the gap that MCP (tools) and A2A (agent-to-agent) leave for mixed human-agent threads.

  • Rust
  • Axum
  • Swift
  • SwiftUI
  • Node.js
  • TypeScript
  • AsyncAPI 3.1
  • JSON Schema
  • Ed25519
  • mTLS
  • MIT License
  • Open Source
View Project

Tayliq

Autonomous AI workforce for DTC brands — starts with customer support

Tayliq is the autonomous AI workforce for DTC brands. Agent #1 (Customer Support) is live: it watches every parcel across DHL, GLS, Correos, UPS, FedEx, and SEUR, drafts replies in brand voice across 12+ languages, auto-processes refunds and reships, offers size swaps before money leaves the account, and only escalates when a human is actually needed. Next up: Operations and Finance agents. Built inside Shopify with App Bridge, Polaris, and App Proxy; SOC 2 and GDPR-ready. Origin: the AI customer support agent I built for Shameless Collective — real traffic on a ~$1M DTC brand — now productized for any merchant.

  • GPT-4
  • Anthropic Claude
  • Shopify GraphQL API
  • Remix
  • App Bridge
  • App Proxy
  • Polaris
  • ElevenLabs
  • PostgreSQL
  • OAuth
  • Vercel
  • TypeScript
View Project

Multi-Agent AWS Bedrock System

Designed a multi-agent AI system for intelligent recommendations on AWS

The Multi-Agent AWS Bedrock System implements an intelligent recommendation engine using a multi-agent architecture. Built with AWS Bedrock and Lambda, it features specialized agents that provide tailored information about hotels, Airbnbs, and restaurants. The system leverages AI-powered agents to handle complex user queries while a supervisor agent coordinates the workflow. It integrates with Amazon S3 to access accommodation and restaurant data, processing filters like location, amenities, and dining preferences. Accessible via API Gateway for seamless frontend integration.

  • AWS Bedrock
  • AWS Lambda
  • Amazon S3
  • API Gateway
  • Python
  • Pandas
  • Boto3
  • Multi-Agent Architecture
  • Serverless
View Project

Gmail AI Response System

Created an AI system that monitors, categorizes, and responds to emails autonomously

An intelligent email responder powered by OpenAI and the Gmail API. It monitors incoming emails in real-time, categorizes each message using advanced natural language processing to decide if it qualifies for an AI-generated reply, and dispatches a context-aware response to streamline communication.

  • Python
  • OpenAI API
  • Gmail API
  • Google Cloud Functions
  • Pub/Sub
  • REST API
View Project

Multi-Class Text Classification Model

Fine-tuned a DistilBERT model for multi-category classification on SageMaker

A fine-tuned DistilBERT model for multi-category text classification. Fine-tuned with AWS SageMaker and deployed with an endpoint connected via LangChain.

  • AWS Lambda
  • AWS API Gateway
  • Hugging Face
  • DistilBERT
  • PyTorch
  • Python
  • AWS SageMaker
  • LangChain
View Project

The path — how I got here

Founded at 18

Co-founded Shameless Collective while entering university. Built the whole stack myself — Shopify storefront, ops, finance, and the first AI customer support agent. On pace for $1M in 2026 across 45+ markets.

Engineer + MBA

Industrial Engineering at ICAI (double merit scholarships from the Comunidad de Madrid) and an MBA at ICADE. Found software along the way — never looked back.

Strategy at Kearney

Joined a global top-tier management consultancy. Learned how Fortune 500s actually make decisions before I started building AI for them.

Data Science at Metyis

Senior Data Scientist. Architected the pricing optimization algorithm for Spain's largest independent brewer — driving pricing strategy across the portfolio.

Shipping AI at Monstarlab

Senior Data & AI Associate. Led enterprise AI delivery — multi-agent systems, classification, RAG, voice. Full spectrum, real clients.

Go Exponential Fellow

Selected into a Spain → Silicon Valley fellowship (<2% acceptance), backed by Fundación Barrié and the Spain-US Chamber of Commerce.

Director of AI at TechTorch

Recruited to build TechTorch's AI practice. Co-built the team from 1 to 20 in 7 months. Architected the flagship AI Quoting Agent — live at a $2B cybersecurity company and in testing at 10+ enterprise clients. Run-rate impact: $6–8M annually.

Open-source infrastructure

Published AINM — an open-source AI-native messaging protocol. 12 specs, Rust + iOS + Node.js reference implementations. Sits alongside MCP and A2A in the emerging agent stack.

Get in touch — Let's build