AI Workflows

I design and ship practical AI workflows that earn their keep. Inputs go in, reliable outputs come out—measured with real funnel metrics. I come from digital marketing, so I care about conversion and ROAS as much as prompts and tokens.

Why work with me

• End-to-end ownership: ingest, normalize, LLM reasoning, paywall/delivery, analytics.
• Guardrails by default: JSON schema, retries, deterministic checks, fallbacks.
• Built for growth: clean UX, fast load, events, attribution, A/B tests.
• Shipped, not theoretical: working code, observability, and clear handover.

Current builds

GTMIntel — AI-Powered Go-To-Market Intelligence

One-liner: Automated platform that monitors 100+ daily startup funding events and transforms them into personalized insights, emails, social posts, and videos for go-to-market teams.

What runs under the hood:

• Data ingestion: Continuous GDELT news monitoring → 15-minute funding event detection.
• AI enrichment: Claude + web scrapers gather company details → OpenAI analyzes needs/challenges.
• Relevance scoring: AI compares startup profiles to subscriber focus → quality scoring.
• Content generation: Automated email briefs, social posts, video segments for each event.
• Personalized delivery: Multi-tier distribution with relevance filtering and growth automation.

Stack & ops: Rust backend, Claude/GPT-4, GDELT feeds, background jobs, email automation, programmatic SEO.

Results: 95% reduction in content creation time, 100+ events processed daily, first-mover advantage for subscribers.

Links: Visit site · Full case study · Join the community

PropertyValuationAI.com — Instant valuation + red-flags (Spain)

One-liner: Paste any Idealista link. Get a price band and Spain-specific risk checks (AFO/LPO, Catastro/Registro mismatches, septic, wells, potencia, etc.) in minutes.

What runs under the hood:

• Scrape & normalize: Apify actor → webhook → cleaned listing facts/images.
• Valuation-lite: €/m² references + heuristics → price band + confidence.
• Risk analysis: LLM with schema-validated outputs and re-prompting on fail.
• Delivery: paywalled preview, checkout via Polar, HTML report + PDF via SES.
• Analytics: PostHog events and attribution (gclid/UTMs) to optimize CVR/ROAS.

Stack & ops: Django/HTMX, Apify webhooks, OpenAI (Responses/JSON), Polar, AWS SES, PostHog.

Links: Visit site · Notes / mini case study · Join the community

How I approach AI workflow builds

Short cycles, clear contracts, and measurable outcomes.

1. Define the job: inputs, outputs, and what “good” looks like.
2. Split logic: deterministic rules first; LLMs for judgment gaps.
3. Guardrail: schemas, retries, test fixtures, and idempotent webhooks.
4. Ship the loop: UX, payments (if needed), delivery, analytics.
5. Optimize: latency, cost per run, and conversion.

FAQs

What tools do you use?

Django/HTMX, Python, Apify/webhooks, OpenAI, PostHog, SES; plus whatever fits the job.

Do you do retainers?

Yes—build + maintain, or advisory on LLM ops and growth experiments.

What’s a typical timeline?

Most MVPs ship in weeks, not months—scoped to a single, valuable workflow.

Have a workflow in mind? Let’s make it real.

Contact me · Book a short call