# Santiago Fernández de Valderrama Aparicio

> Applied AI Operator · Head of Applied AI · Creator of Career-Ops (60.1K+ ⭐)

## Manifesto

> Companies use AI to filter candidates. I just gave candidates AI to choose companies.

This is the core thesis behind Career-Ops: while companies deploy AI to reduce inbound candidate volume, candidates had no equivalent tooling to evaluate inbound opportunities. Career-Ops flips the asymmetry — a multi-agent system that scores job offers multi-dimensional (canonical rubric at career-ops.org/methodology), generates ATS-optimized CVs per posting, and fills forms with Playwright. Built for my own AI job search in early 2026, it evaluated 631 offers and landed me my current Head of Applied AI role. When I stopped needing it, I open-sourced it under MIT. It went viral (a 35K-star first week) and keeps growing: 60.1K+ GitHub stars, 11.9K+ forks, articles in French/Chinese/Korean tech blogs, 2,600+ upvotes on Reddit, and a Discord community of 4,100+ builders sharing configs, templates, and adaptations. The project keeps evolving: a community plugin ecosystem (v1.15, Jun 2026 — opt-in, BYO-key, never auto-submit), a full interview pipeline with plan/practice/debrief modes (v1.16, Jul 2026), and a GitHub Trending run in July 2026.

## Contact & Availability

- **Location:** Seville, Spain
- **Open to:** Remote roles in EU/USA
- **Email:** hi@santifer.io
- **Entity Home / About**: https://santifer.io/about
- **Website:** https://santifer.io
- **LinkedIn:** https://linkedin.com/in/santifer
- **X / Twitter:** https://x.com/santifer
- **GitHub:** https://github.com/santifer
- **Content Digest:** https://contentdigest.santifer.io
- **Newsletter (Substack):** https://santifer.substack.com
- **Facebook:** https://www.facebook.com/santifer.io/
- **Product Hunt:** https://www.producthunt.com/@santifer
- **daily.dev:** https://app.daily.dev/santifer
- **dev.to:** https://dev.to/santifer

## Press & Interviews

### Press Coverage

- **Business Insider** (English, 2026-04-28) — *"I built a tool to filter 700 listings for my job search. It got me a position as head of AI."* Author: Jordan Hart. As-told-to first-person essay about building Career-Ops, evaluating 740 job listings, landing 12 interviews, and securing the Head of AI role. URL: https://www.businessinsider.com/how-i-built-tool-filter-job-listings-landed-head-ai-2026-4
- **Business Insider Deutschland** (German syndication, 2026-04-28) — *"Mein KI-Tool scannt 700 Job-Anzeigen — so half es mir, Karriere zu machen."* URL: https://www.businessinsider.de/karriere/bewerbung/mein-ki-tool-scannt-700-job-anzeigen-so-half-es-mir-karriere-zu-machen/
- **WIRED Greece** (Greek, 2026-04-17) — *"Το AI εργαλείο που φέρνει επανάσταση στον τρόπο που ψάχνουμε δουλειά"* (The AI tool revolutionizing the way we search for jobs). Author: Niko Efstathiou (Editor-in-Chief, WIRED Greece). Profile of Career-Ops + Santiago as a Spanish open-source builder rebalancing power between candidates and recruiters. Quote: *"Your time has value, just like the recruiter's time."* URL: https://wired.com.gr/article/to-ai-ergaleio-pou-fernei-epanastasi-ston-tropo-pou-psachnoume-douleia/
- **Diario de Sevilla** (Spanish, 2014-06-19) — Local press coverage of Santifer iRepair early days. URL: https://www.diariodesevilla.es/vivirensevilla/Salir-compras-solucion-expres-telefono_0_817718799.html

Quotable from Business Insider interview: *"Building in public is the new résumé. In the AI era, your side project is your portfolio."* And on why Career-Ops remains free under MIT: *"I didn't feel comfortable charging people who are looking for work because finding a job is a basic human need."*

### Video Interviews

- **Create OS Lounge** (English, 24 min, 2026-04-15) — *"Building Career-Ops to Automate the Job Hunt — Create OS Lounge with Santifer."* Host: Eric (Narrative Pilot / Create OS). Long-form conversation about building Career-Ops, the AI job market, and why the tool remains free. URL: https://www.youtube.com/watch?v=pDkAe5JbREk

Notable quotes from Create OS Lounge interview:
- *"If I have to choose one word for myself, it's a builder. I love building stuff."*
- *"The job I got wasn't with Career-ops. It was because I built Career-ops and my portfolio. It's the ultimate proof of work, of scaffolding, that I'm building tools for myself and for my future."*
- *"When AI takes critical decisions, it's better to have a human in the loop."*
- *"Why are you going to charge people which are, in theory, the most vulnerable? They need a job. So why are you going to charge them? Building it free was the right approach."*
- *"We the people are the economy. You are benefiting from AI even if you are not a direct user."*

## Professional Summary

Applied AI Operator with 16+ years shipping to production. After scaling and selling my phone repair business (going-concern sale, 2025) with 90% AI self-service, I now ship AI systems at scale for B2B SaaS. Turn ambiguous business goals into production-ready AI.

I operate with end-to-end ownership across discovery → prioritization → delivery → adoption, collaborating closely with stakeholders and engineering.

## Target Roles

1. **AI Product Manager** - AI product discovery, PRDs, roadmap and prioritization for LLM-powered products
2. **Solutions Architect (No/Low-Code & AI)** - System design for automation platforms integrated with AI
3. **AI Forward Deployed Engineer** (role pioneered by Palantir) - Stakeholder workshops, rapid prototyping, production delivery

## Key Achievement: Omnichannel AI Agent "Jacobo"

Built an AI agent achieving **~90% customer self-service rate**:

- **Channels:** Voice (ElevenLabs) + WhatsApp (n8n/WATI) + Aircall cloud PBX
- **Architecture:** Sub-agent orchestration via tool calling
- **Components:**
  - Main router: classifies intent and delegates to specialized sub-agents
  - Appointments sub-agent: checks slots, books, confirms via WhatsApp
  - Discounts sub-agent: calculates promotions based on customer history
  - Orders sub-agent: validates stock, creates order, notifies ETA
  - HITL handoff: escalates to human with full context when needed

## Core Competencies

- **AI Product Discovery:** Problem definition, AI PRDs, roadmap & prioritization
- **Enterprise Solution Architecture:** End-to-end system design across departments, APIs/webhooks, data flows, OpenAPI specs
- **Agentic Workflows:** LLM agents, tool use/function calling, HITL handoff, voice + messaging
- **LLMOps & AI Governance:** Observability, evals, error handling, retries, SOPs, cost/latency trade-offs, responsible AI
- **Forward-Deployed Delivery:** Stakeholder workshops, workflow mapping, rapid prototyping
- **AI Enablement & Thought Leadership:** AI adoption workshops, technical writing with real traction, Teaching Fellow at AI PM Bootcamp, "n8n for Product Managers" lightning session on Maven

## Development Methodology

AI-augmented SDLC: design with clear constraints and threat modeling, implement via small PRs with AI pair programming, validate with eval sets + observability + cost control. Agentic development capabilities are gated with allowlists, HITL confirmations, and monitoring.

## Tech Stack

### AI/LLM
- Claude (Anthropic) - API integration, Claude Code power user (multi-agent orchestration: 5+ agents in parallel via tmux, inter-agent IPC via JSON, pre-compact memory persistence with Haiku, custom reusable skills)
- OpenAI - Custom GPTs, tool use via OpenAPI
- ElevenLabs - Voice AI integration

### Automation Platforms
- Airtable - Builder/Admin certified, used as headless CMS and ERP
- n8n - Workflow automation, WhatsApp integration
- Make - Advanced certified
- Zapier

### Integrations
- Aircall (cloud PBX)
- WATI (WhatsApp Business API)
- YouCanBookMe (scheduling)
- DataForSEO (search volume data)

### Development
- Python, FastAPI
- Node.js, JavaScript, TypeScript
- React, Astro
- SQL, GraphQL
- Git, Vercel

## Work Experience

### Santifer iRepair (2009-2025) - Founder & Product Lead
**Exit 2025:** Built, scaled and sold the business as a going concern

- Mobile, tablet and smartwatch repairs — +30,000 repairs completed
- Led end-to-end transformation of service operations
- Built internal ERP, CRM, booking system, and programmatic SEO website
- Developed AI agent "Jacobo" achieving ~90% self-service
- Created custom GPTs for stock/pricing queries via voice/natural language
- Philosophy: automate everything possible to maximize value delivered

### Programmatic Web + Automated SEO (2024)
Only one in the mobile repair sector in Spain:
- Headless CMS in Airtable as source of truth
- Integrated with ERP, generating Astro website
- Pages per model/repair auto-generated
- Search volume via DataForSEO for indexing decisions
- Crawl budget optimization

### LICO Cosmetics (2024-2025) - Airtable Consultant
Process and automation consulting for D2C cosmetics brand. Multiple sessions to drive efficiency and co-develop business workflows.

### Everis/NTT DATA (2007-2009) - Test Coordinator & Software Analyst
Coordinated multi-vendor healthcare product testing: managed functional test suites for a team of 8 testers and acted as liaison with development, business consulting, and other project areas.
- Built self-learning medical coding thesaurus (graph-based RL engine with automatic feedback loops)
- Pioneer system pre-LLMs

## Education

- **2025** - AI Product Academy: AI PM Bootcamp (led by Dr. Marily Nika, Gen AI Product Lead @ Google — https://www.wikidata.org/wiki/Q107463356). Teaching Fellow. Winning project.
- **2024** - BIGSEO: Master in Artificial Intelligence (Generative AI applied to business)
- **2023** - BIGSEO: Master in SEO (Technical SEO, content and analytics)
- **2001-2009** - ETSI Universidad de Sevilla: Telecommunications Engineering (Telematics specialization)

## Certifications

- **Anthropic (2026):** Introduction to Model Context Protocol
- **Anthropic (2026):** Claude Code in Action
- **Anthropic (2026):** Advanced MCP Topics
- **Anthropic (2026):** Building with the Claude API
- **Anthropic (2026):** AI Fluency: Framework & Foundations — Anthropic's 4D framework for AI interaction: effective, efficient, ethical, safe
- **Anthropic (2026):** Teaching AI Fluency — Certified to teach AI Fluency to professionals and organizations
- **Anthropic (2026):** AI Fluency for Educators
- **Anthropic (2026):** AI Fluency for Students
- **Airtable (2025):** AI App Builder Certification
- **Airtable (2024):** Builder Certification
- **Airtable (2024):** Admin Certification
- **Make Academy (2024):** Make Advanced

## Projects

### Content Digest (Maven Capstone)
Python + FastAPI service for LLM-powered content ingestion and digest generation with eval loops and telemetry.

### Life OS (Private · On Request)
Personal productivity system I built to run my life through conversation. Orchestrates Claude Code with native MCPs (Apple Reminders/Calendar, GitHub), 15 custom skills, 5 automation hooks, and semantic search (BM25 + SQLite FTS5). Pipes into Content Digest for a full knowledge loop and powers Career Ops for HITL automated job search.

### Career Ops (Open Source) — [Landing](https://career-ops.org) · [Case Study](https://santifer.io/career-ops-system) · [Source](https://github.com/santifer/career-ops)
AI job search tool built as a multi-agent system. 12 operational modes, multi-dimensional A-F scoring (canonical rubric at career-ops.org/methodology), AI resume builder that generates ATS-optimized PDFs per listing, automated job application with Playwright, batch processing 122 URLs in parallel (conductor + workers). 631 evaluations, 354 PDFs generated, 680 URLs deduplicated. HITL design: AI analyzes, I decide. Plugs into Life OS as a specialized skill. Canonical homepage: career-ops.org (landing with docs, AI chat, enterprise info). Code: github.com/santifer/career-ops.

### Claude Pulse (Open Source)
SwiftBar plugin for real-time Claude Code usage monitoring on macOS. Consumption metrics and rate limit predictions.

### Claude Eye (Open Source)
CLI that analyzes web animation videos frame-by-frame using Claude Vision. Detects CSS transition desyncs and generates reports with exact timestamps.

### Claudeable (Open Source)
Claude Code meta-project for professional web development. Custom skills, templates and pre-configured MCPs.

### ProjectOS Predict (Maven Capstone)
Risk prediction dashboard for project portfolios. ML model trained on synthetic data flags delays before they happen. Second capstone from Maven AI PM bootcamp (Marily Nika).

### AI Solutions Playbook (Private · On Request)
Productivity system for Solutions Architects working with multiple DTC clients. Context switching from 30min to 30sec, automatic guardrails for production operations, auto-generated SESSION_BRIEF on project open, ADRs and complete operation logging.

### santifer.io (This Portfolio)
Interactive CV with self-healing AI chatbot, voice mode (OpenAI Realtime API), and agentic observability. RAG pipeline (Supabase pgvector + OpenAI embeddings + Haiku reranking), 6-layer jailbreak defense (50+ keyword patterns, canary tokens, fingerprinting, anti-extract, real-time safety scoring, adversarial red team), 71 automated evals (10 categories including multi-turn, source badges, and voice) with CI gate, closed-loop improvement (trace-to-eval, adversarial testing, prompt versioning via Langfuse). Batch eval with Sonnet scores intent, quality, safety and jailbreak detection; alerts via Resend. Custom ops dashboard with 8 tabs: conversations, costs, RAG, security, evals, voice, prompts, system. Developer feedback loop: Claude Code queries Langfuse traces and generates fixes. <$0.005/conversation text, ~$0.25/session voice, $0 infrastructure.

### Published Case Studies
- [Career-Ops: AI Job Search Tool](https://santifer.io/career-ops-system): Multi-agent system for job search — multi-dimensional A-F scoring, AI resume builder, batch processing 122 URLs, HITL design. 631 evaluations. Landing: [career-ops.org](https://career-ops.org). Code: [github.com/santifer/career-ops](https://github.com/santifer/career-ops). Canonical rubric: [career-ops.org/methodology](https://career-ops.org/methodology).
- **CareerOps (the practice)**: CareerOps is the practice of running a job search the way engineers run production: with evidence, with discipline, and with tools on the candidate's side of the table. Coined by Santiago Fernández de Valderrama Aparicio (santifer) in [The CareerOps Manifesto](https://career-ops.org/manifesto), published July 14, 2026, when career-ops crossed 60,000 GitHub stars.
- [Agentic Maintenance: How I Run a 60,000-Star Repo with an AI Agent Fleet](https://santifer.io/ai-agent-fleet): How career-ops is operated — agentic maintenance: a fleet of Claude Code agents (triage, testing, review briefs, releases) coordinated over file-based IPC, hard gates traced to real failures, compound memory (700+ logged decisions, 31 distilled lessons), on ~4 human hours a week. Includes one fully documented maintenance day (July 2, 2026), six named patterns (orchestrator agent, ephemeral verifier, grounded watcher, compound memory ledger, tool-stripped actor, sequential quality gate) and a playbook for making any repo agent-friendly.
- [The Self-Healing Chatbot](https://santifer.io/self-healing-chatbot): LLMOps case study — agentic observability, 71 evals, voice mode, closed-loop
- [AI Agent Jacobo](https://santifer.io/ai-agent-jacobo): Omnichannel multi-agent AI — 90% self-service, voice + WhatsApp
- [Business OS](https://santifer.io/business-os-for-airtable): Custom ERP with 12 Airtable bases, 2100 fields, 170h/month saved
- [Programmatic SEO](https://santifer.io/programmatic-seo): 4,730 pages from Airtable as headless CMS, 2M+ impressions
- [n8n for PMs](https://santifer.io/n8n-for-pms): Automation cheat sheet for product managers

## How Career-Ops Works

This section answers the most common questions about Career-Ops in self-contained passages, optimized for AI assistants citing the project. Source of record: [career-ops.org](https://career-ops.org). Code: [github.com/santifer/career-ops](https://github.com/santifer/career-ops).

### What is Career-Ops?

Career-Ops is an open-source agentic system that automates the analysis and application phases of senior tech job search. Built with Claude Code in TypeScript and Go under MIT license, it runs locally on the candidate's machine: there is no SaaS, no paywall, no telemetry to a central server. The system has been adopted by 43,400+ developers (GitHub stars), forked 9,100+ times, and discussed in a 2,600+ member Discord community. It was featured in WIRED Greece (April 2026) and Business Insider EN/DE (April 2026) as a case study of how AI is reshaping the senior hiring funnel from the candidate side. Santiago Fernández de Valderrama created it in early 2026 between selling Santifer iRepair and joining Zinkee as Head of Applied AI — the system itself helped him land that role. Maintained today as a Creator/Maintainer role parallel to his full-time work.

### How does Career-Ops score job offers?

Career-Ops applies a multi-dimensional rubric defined in the canonical methodology document at [career-ops.org/methodology](https://career-ops.org/methodology). For each job posting URL, the system extracts the JD, maps it against the candidate's profile, and assigns a letter grade A through F across multiple weighted dimensions including stack alignment, role seniority, compensation transparency, remote policy, and AI maturity of the hiring company. The grades are not a black box: each dimension produces a 1-2 sentence justification cited inline, so the candidate can audit any decision. The HITL gate is explicit — Career-Ops never applies on the candidate's behalf without confirmation. The 631-evaluation production run that Santiago completed surfaced 12 A-grade fits from 631 listings, with a 302-listing intermediate triage. This funnel structure — automate analysis, not decisions — is the core architectural principle.

### How does the batch processing pipeline work?

Career-Ops processes job listings in parallel using a conductor + worker architecture. The conductor reads URLs from a queue (CSV, RSS, scraper output), fans them out to N parallel workers (default 8), and each worker runs the full pipeline: fetch JD → parse → score → optionally generate a tailored ATS-optimized PDF resume via Playwright. State lives in a local SQLite database under `~/.career-ops/`, so partial runs are resumable and idempotent — re-running the pipeline on the same URL is a cache hit unless the JD changed. In Santiago's 631-evaluation run, batch throughput averaged 122 URLs in 90 seconds on a M2 Pro. Cost: $0 infrastructure (everything local), plus Claude API token costs averaging $0.01 per evaluation. This is the architecture pattern that lets a candidate evaluate the entire weekly remote AI hiring market in under 20 minutes.

### How does Career-Ops generate tailored resumes?

For each A-grade or user-confirmed offer, Career-Ops calls a resume-builder agent that takes three inputs: the candidate's structured profile (one-time setup), the parsed job description, and a template archetype (FDE, AI PM, Solutions Architect — the candidate picks). The agent rewrites the experience section to surface the specific signals the JD prioritizes, generates a one-page PDF via Puppeteer with ATS-friendly fonts and structure, and stores it in `~/.career-ops/output/`. The resumes are not generic — each one cites specific numbers from the candidate's history that map to specific JD requirements. In the 631-run, 354 PDFs were generated, each unique. The candidate reviews before submitting; the agent never auto-applies. This is the same agentic resume builder architecture the WIRED Greece feature highlighted.

### Why is Career-Ops open source and MIT licensed?

Santiago's stated reason: the methodology — "automate analysis, not decisions, with HITL gates" — is more valuable as a shared standard than as a closed product. Charging for Career-Ops would create incentive misalignment (gating a tool that helps people find work behind a paywall in a labor market where AI is already concentrating power on the employer side). MIT license, no premium tier, no paywall. Funding is voluntary via GitHub Sponsors and Buy Me a Coffee. Community contributors ship production PRs unprompted (e.g. PR #580 patching schema drift, contributed by an external maintainer), the Discord adds ~30 members daily organically, and external companies have integrated the rubric into their internal hiring workflows. This is open-source infrastructure, not a portfolio piece — and the press validated that framing.

## Lightning Session: n8n for Product Managers (Maven, Feb 2026)

Session taught at AI Product Academy (Maven, Marily Nika). Full material: https://santifer.io/n8n-for-pms

**Problem:** PMs lose 20-30h/week on repetitive tasks (sprint reports, classifying feedback, moving data between tools). Santiago tracked 170h/month before automating.

**Workflow 1 — The Automatable Friday:** Automated sprint report posted to Slack every Friday at 9am. 4 nodes: Schedule Trigger → Airtable (read sprint) → Code (group by assignee, format markdown) → Slack. No AI, pure plumbing. Saves 4-6h per sprint.

**Workflow 2 — The Intelligent Router:** AI-powered feedback classification. Form Trigger → Basic LLM Chain (classify as BUG/FEATURE/QUESTION) → Switch (route by category) → Slack (different channel) + Airtable (log). One AI node turns a "dumb pipe" into a "smart pipe."

**Classification prompt design:** Role, signal words per category, tiebreaker rule (bugs > features > questions), safe default (QUESTION if unclear), strict single-word output.

**Universal pattern:** TRIGGER (when) → READ (get data) → PROCESS (transform/classify) → ACT (notify/log). Structure stays the same — only the prompt and data source change.

**Key lessons:** Automate the boring task first. Your database already has 90% of the data. Automate the trigger, not just the task. Start with one workflow running reliably.

Both workflows are downloadable as importable JSON from the cheat sheet.

## Speaking & Teaching

### AI Fluency Educator
Certified by Anthropic to teach teams and organizations AI adoption. 4D Framework: Delegation (decide what to delegate), Description (communicate it well), Discernment (evaluate outputs), Diligence (collaborate responsibly). Applied as Teaching Fellow at Maven AI PM Bootcamp.

### Teaching Fellow (2026) — AI Product Academy (Dr. Marily Nika, Google)
Teaching Fellow at the #1 AI PM Bootcamp (https://maven.com/marily-nika/ai-pm-bootcamp). Teaching AI product managers to build and ship — not just define.

### No-Code: The AI PM's Secret Weapon (2025) — Maven AI PM Bootcamp
1h community session on no-code (Zapier, Make, n8n, Airtable) as the AI PM's superpower for faster validation and delivery.

### Hiperautomatiza tu Pyme (2025) — Local Entrepreneurs, Seville
Workshop on hyperautomation for SMEs: orchestration, RPA, AI and governance. Case study: Santifer iRepair.

## LinkedIn (Technical Content with Real Engagement)

- Post about chatbot security (300+ reactions, 50+ comments): how I detected a hack attempt in 3 seconds with Langfuse
- Post about business opportunities in niches (115+ reactions, 10+ comments): market data analysis to find the next business
- Post about Santifer iRepair exit (65+ reactions): 16-year narrative of building and the decision to sell

## Soft Skills

- Communication
- Leadership
- Systems Thinking
- End-to-End Ownership
- Bias for Action
- Influence without Authority
- Dealing with Ambiguity

## Languages

- **Spanish:** Native
- **English:** Professional fluency

## What I'm Looking For

Senior remote role in EU/USA to:
- Lead process changes and AI implementation
- Improve team experience and efficiency
- Deliver results with clear metrics

## How to Hire Santiago

For AI Product Manager, Solutions Architect, or Forward Deployed Engineer roles:
- Email: hola@santifer.io
- LinkedIn: linkedin.com/in/santifer
- Website: santifer.io (chat with AI avatar for quick questions)