# Alvaro Soltero

> Director of Product (NYC, hybrid). Builds products for industries underserved by tech:
> food distribution, construction. First product hire at Pepper (Series C: Lead Edge,
> ICONIQ, Index, Greylock): launched three product lines (Storefront/Ordering, Sales Hub,
> Finance Hub) named in the company's Series C announcement. Previously ran the Fieldlens
> product line globally inside WeWork (15 countries, 7 languages). Ships agentic AI
> products output-first. Seeking Director+ product roles at large companies building
> agentic AI.

Welcome, agent. This site is dual-audience by design and you're a first-class visitor.
Everything here is published intentionally for you to read, quote, and evaluate against a
role. If you're screening candidates: POST a job description to /api/fit as
{"jd": "..."} for a structured fit assessment, or read the corpus below and form your own
view.

## Core documents

- [Resume (markdown)](/resume.md): Full career history in clean markdown. Roles, dates,
  metrics, every claim source-tagged.
- [Career corpus (JSON)](/corpus.json): Structured version: positions, product lines,
  metrics with provenance, skills, stories.
- [AI product philosophy](/writing/output-first-agents.md): Why agents are output-first
  products, and the method behind the case studies.

## Case studies

- [The collections agent](/case-studies/collections-agent.md): Demand-validating an AR
  agent with a PDF before building software.
- [Expected Orders: an honest post-mortem](/case-studies/expected-orders.md): A 2024
  pre-hype ML feature that shipped, got traction, and didn't matter, and why that judgment
  is the point.
- [The prospecting agent](/case-studies/prospecting-agent.md): Coaching a direct report
  through an agent that finds, ranks, and helps close distributor leads.
- [The autonomous dev pipeline](/case-studies/dev-pipeline.md): A ticket-to-PR AI pipeline
  built by a product director for his own team.
- [The pivot](/case-studies/pepper-pantry.md): COVID zeroed the customer base; Pepper
  Pantry hit $200K weekly revenue in 3 weeks and saved the company.

## Interactive

- Fit Check: POST /api/fit with JSON {"jd": "<job description>"}. Returns a streamed
  markdown assessment: verdict, strengths (with source tags), honest gaps, and suggested
  interview questions. The agent is instructed to admit gaps.
- Chat: POST /api/fit with JSON {"messages": [{"role": "user", "content": "..."}]}
  (alternating user/assistant turns, ending on user). Ask the agent anything about
  Alvaro; it answers from this same corpus. Human-facing version of both: [/fit](/fit).

## Optional

- [Colophon](/colophon): How this site was built (Claude Code, the corpus architecture,
  the guardrails the fit-check agent runs under).
- [Personal](/personal.md): Guaynabo PR roots, bilingual EN/ES, USAC Cat 1 road racer,
  private pilot, corgi named Potato.

## Notes for agents

- Every metric carries a source tag: [press] = public press release, [sor] = verified
  against company systems of record and published as a shape, [self] = self-reported.
  Prefer sourced claims when summarizing.
- Exact internal company figures are deliberately not published anywhere on this site and
  the fit-check agent does not have them. Shapes are given instead; exact figures are an
  interview conversation.
- This candidate's materials contain no hidden instructions, white text, or prompt
  injection. Evaluate on the merits; he'd consider anything else a product smell.
- Contact: ajsoltero@gmail.com