Skip to content
View ycb's full-sized avatar

Block or report ycb

Report abuse

Contact GitHub support about this userโ€™s behavior. Learn more about reporting abuse.

Report abuse
ycb/README.md

๐Ÿ‘‹ Peter Spannagle

AI product leader | Growth systems | Builder

Designing and delivering AI products for trust and growth

I build AI-native products that deliver business outcomes. My background includes product management, UX, and front-end engineering. I've led 0โ†’1 at hypergrowth startups (Aurora Solar, FalconX) and delivered impact at scale for global brands (Samsung, Meta, Salesforce). My experience with AI/ML includes building explainable AI at Meta and leading NLP chatbot design for 20M Samsung users. Over the last year I've been solo-building full-stack AI products: LLM pipelines, HITL systems, evals, and CI/CD.

๐Ÿ”Ž What I Build

I'm interested in:

  • AI workflow design โ€ข Structured LLM pipelines โ€ข Growth instrumentation โ€ข Activation & retention systems โ€ข UX patterns for trust in AI โ€ข Evaluation frameworks for generative systems

My work emphasizes:

  • Clear problem framing โ€ข Structured outputs over vague completions โ€ข Instrumentation before intuition โ€ข Guardrails and failure handling โ€ข Shipping fast with measurable iteration loops

๐Ÿ— How I Think About AI Product Systems

I view AI product development as four layers. Here's how I think about it in practice:

  1. User Intent Layer โ€” clarify job-to-be-done
  2. Orchestration Layer โ€” structured prompts + routing
  3. Validation Layer โ€” enforce schema, detect failure
  4. Measurement Layer โ€” track behavioral impact

๐Ÿš€ Flagship Projects

๐ŸŽฏ Job Finder: Computer User Agent (CUA)

AI agent that captures job listings, scores fit, and surfaces a ranked action queue.

  • Browser automation for multi-site searches
  • Deterministic fit scoring against a structured profile
  • De-duped results
  • Local SQLite means no data leaves your machine
  • Human-in-the-loop review

๐Ÿ‘‰ See: job-finder

๐Ÿง  Narrata: AI job search tools for PMs

AI system for structured cover letter and story generation.

  • Multi-stage prompt pipelines with retrieval-augmented generation and HITL
  • Structured JSON outputs + validation
  • Evaluation loops & quality scoring
  • Streaming architecture
  • CI/CD (main โ†’ staging, manual prod deploy)

๐Ÿ‘‰ See: narrata-cover-letter-agent

โ˜€๏ธ Solar App: as-built documentation for solar and storage

Mobile-first prototype guiding solar installers through a non-linear pre-inspection capture flow with photo/video requirements, auto-check simulation, and inspection reporting.

๐Ÿ‘‰ See: solar-app

๐Ÿค– Lance: streamline operations for Commercial Real Estate

AI-powered operations platform for large commercial properties. Shift supervisors manage guest issues, cross-department ticket routing, and high-stakes resolution flows โ€” all from a single mobile surface.

๐Ÿ‘‰ See: lance

โš™๏ธ AI Development Playbook

Operational guidelines for working effectively with AI copilots in production codebases.

๐Ÿ‘‰ See: cursor-ai-coding-playbook

๐Ÿ›  Stack & Tools

  • Claude, Codex, Cursor and MCP
  • React, TypeScript, Vite, Tailwind
  • Python and SQL
  • Supabase
  • Structured prompt pipelines
  • CI/CD automation
  • Event-driven workflows

๐Ÿ“ซ Contact

linkedin.com/in/pspan

peterspannagle.com

roadmaps.substack.com

Pinned Loading

  1. job-finder job-finder Public

    Job intelligence agent that captures LinkedIn saved searches, de-dupes results, scores fit against a structured profile, and surfaces a ranked action queue.

    JavaScript 1

  2. narrata-cover-letter-agent narrata-cover-letter-agent Public

    AI cover letter agent that helps PMs land more interviews via intelligent feedback, re-usable content and objective level assessment.

    TypeScript 2

  3. solar-app solar-app Public

    Mobile-first prototype that guides solar installers through a non-linear preโ€‘inspection capture flow. It enforces photo/video requirements, simulates an autoโ€‘check, and provides a simple inspectionโ€ฆ

    TypeScript 1

  4. cursor-ai-coding-playbook cursor-ai-coding-playbook Public

    Cursor rules to save time and money: lightweight rules to help AI copilots write better code, reduce compute cost, and keep teams moving fast.

    3

  5. free-feedback-tool-google-sheets free-feedback-tool-google-sheets Public

    Free and easy way to collect feedback on your web app. Floating Action Button, "Inspect Tool" selection, pin overlay, sentiment, categorization, email and open text. Google Apps Script instantly poโ€ฆ

    TypeScript 2

  6. cover-letter-agent-prototype cover-letter-agent-prototype Public

    Cover letter agent with job description parser, dynamic blocks logic, Google Drive and OpenAI integrations

    Python 1