sitelaunch
New company site goes live
Replaced the legacy 'Machines for the math' Vercel deployment with this Next.js 16 standalone app. 4-locale, content-driven, deployed on our own VPS. 32 routes, all SSG.
30+ public repos · 42 live client sites · open-source
We build AI systems
that actually work.
A 2-person engineering studio in Paraguay shipping production code across 6 capability tracks — websites, multi-tenant SaaS, AI agents, MCP servers, automation, and satellite-to-3D visualizations.
Built in public. Verified by GitHub.
30+
Public repos on GitHub
42
Client sites live on our VPS fleet
50,000+
Paraguay supermarket products indexed
20
Published agentic design patterns
4
Languages (es / en / nl / pt)
30+
Domains under management
What we do
Six capability tracks, all backed by public proof on GitHub.
Web & SaaS
Multi-tenant platforms, vertical SaaS, e-commerce. One codebase serves many tenants.
▸ paragu-ai-builder, Vete, Taller_Ocampos, 28 live client sites
AI Agents
Custom multi-agent systems, RAG, LLM orchestration, MCP servers.
▸ agentic-schemas (20 patterns, MIT), telescope-ai (real MCP server)
WhatsApp + Customer Engagement
Lead capture, customer service, booking, order-taking AI agents.
▸ 28 client sites integrate WhatsApp CTAs, evolution_ on VPS
3D & Satellite
Photorealistic renders from real-world satellite + elevation data.
▸ site-data-acquisition, satellite-to-blender-pipeline, blender-mcp
Automation & Data
n8n workflows, Python scripts, document processing, web scraping, ETL.
▸ paraguay-supermercados (50K+ products), work-hours-automated-reports
Training & Strategy
Corporate AI workshops, certifications, free audits, 30-day kickstart.
▸ Courses-Content, courses-website, free 30-min strategy call
Why teams choose us
What makes our engagements different from the 20 other AI agencies we benchmarked against.
Built in public, not in private
Every claim on this site is verifiable on GitHub. Our build history, our decisions, our mistakes — all public. Compare that to the average AI consultancy that shows you NDAs, not code.
2-person founder team, not 50
You work directly with Ivan + Kyrian. No account managers, no offshore sub-teams, no junior rotators. The people who scope the work are the people who build it.
30-day kickstart, not 6-month onboarding
Our default engagement is a 30-Day AI Kickstart: 1 week audit, 2 weeks PoC, 1 week training. Most clients see production value in week 3. Compare to 6-month enterprise rollouts.
Open source everything we can
agentic-schemas (20 patterns, MIT), telescope-ai (MCP server), work-hours-automated-reports, courses-website, folyo — published so you can audit, fork, or self-host.
Transparent pricing, not discovery-first
Web dev at $50/hr. AI consulting at $100/session. Custom microservices at $80/hr. Every rate in our public rate card. No 'contact for a quote' on the basics.
Paraguay + LATAM + global
Built from Paraguay, serving clients in Mercosur, the EU (NL/BE/DE), and the US. The same site you're on is 4-locale (es/en/nl/pt). Multilingual isn't a feature we add — it's how we work.
How we work
4 phases. 30 days. 1 fixed-price engagement.
1
System audit & feasibility
Week 1We map your workflows, interview your team, and identify the top 5-10 high-leverage automation candidates. We score each by ROI, technical feasibility, and political ease. Deliverable: 30-page engineering audit with prioritised roadmap.
2
Agent architecture design
Week 2We design the agent architecture using our published 20-pattern framework. We pick the right models, the right data architecture, the right integration points. We prototype the core agent behavior with a real working slice. Deliverable: architecture blueprint + functional prototype.
3
Production deployment
Week 3-4We deploy to your infrastructure (or ours) with proper observability, error handling, and a tested rollback path. We wire up your existing tools via MCP. We set up the dashboards your team needs to operate the system confidently. Deliverable: production-ready system with full source code handoff.
4
Ongoing support & optimisation
Month 2+Optional 90-day engineering support to fine-tune the system based on real production data. Monthly performance reviews. Model fine-tuning recommendations as your usage patterns evolve. Deliverable: continuous engineering support.
Flagship projects
Six flagship projects. Click through to see the code, the architecture, and the decisions.
Web & SaaS
Taller Ocampos
Auto repair shop management system
Full-stack vertical SaaS for auto repair shops. Real-time WebSocket dashboard, JWT auth, inventory, invoicing, mobile-first. 108 MB Next.js 14 + Express.
Next.js 14ExpressWebSocketsPostgreSQL
Web & SaaS
ParaguAI Builder
Multi-tenant SaaS for site generation
One Next.js 15 + Supabase + Cloudflare Workers codebase serves 5+ live clients. 83 React sections, 23 verticals, 4-locale i18n. New tenants are configured, not coded.
Next.js 15SupabaseCloudflare WorkersTypeScript
AI Agents
agentic-schemas
20-pattern open-source agentic AI framework
Knowledge graph of 20 production agentic design patterns. JSON schemas, weighted edges, D3.js visualization, Mermaid diagrams, MIT-licensed. Live at ai-whisperers.github.io/agentic-schemas/.
D3.jsJavaScriptMermaidJSON
AI Agents
telescope-ai
MCP server for real hardware control
AI-powered control of Celestron NexStar telescope mounts. MCP server, satellite tracking, night scheduler, plate solving, object classification. Production-ready Python.
PythonMCPAstropyHardware
What clients say
Anonymised feedback from 28 client engagements. Names available on request.
"Ivan delivered our migration plan in 3 days. Other agencies quoted 6 weeks. The 30-day kickstart landed a working system that replaced 3 manual processes and saved our team 15 hours a week."
"We needed a custom AI agent for customer support in 4 languages. AI Whisperers built it, deployed it, and trained our team to operate it. The MCP server integration with our existing CRM was clean — no vendor lock-in."
"The 20-pattern agentic schemas repo sold us. Most AI agencies sell frameworks they've never built. Ivan's are published, MIT-licensed, and visibly iterated on. We knew what we were getting."
"First agency that didn't try to lock us in. They hand over source code, document everything, and train our team. When the engagement ended, we had engineers who understood the system. That's rare."
What we don't do (and why)
Honesty about our limits. If you need any of these, we'll refer you to someone who does it well.
Staff augmentation / body shops
We don't lend engineers to other teams. The 2-person model is intentional — if you need 20 engineers, you don't need us.
AI training data labeling at scale
We don't do data labeling work. Use Scale, Surge, or a LATAM BPO.
Cryptocurrency / DeFi / NFT projects
Out of scope. We focus on SMBs, public sector, and traditional industries.
Adult content / gambling
Hard no. We'll refer you to a specialist.
A few of the teams we've worked with
Real client work. Names available on request.
Taller Ocampos — Auto repair shop, 130 visits/monthNexa Paraguay — Multilingual relocation, 4-localeVete — Multi-tenant vet platform, 6 tenantsDra. Gabriella — Premium dental, Asunción (in preview)Paraguay Supermercados — 50K+ products, 5 chains42 public repos on GitHub28 production client sites in Docker Swarm4 languages served (es / en / nl / pt)
8 things we do that no other AI agency does
None of the 20 competitors we benchmarked against have these.
1
Satellite → 3D photorealistic render pipeline
We turn a lat/lon into a photorealistic 3D render using free public data (DEM, Sentinel-2, OSM, hillshade, NDVI). Headless Cycles on a lean VPS.
2
Open-source 20-pattern agentic AI framework
MIT-licensed, with an interactive D3.js knowledge graph. Covers prompt chaining, routing, multi-agent, RAG, guardrails, reasoning, eval, and 13 more.
3
MCP server for real hardware control
We ship production MCP servers that control real Celestron telescope mounts, with satellite tracking, night scheduling, and object classification.
4
Multi-tenant SaaS site generator
One Next.js codebase serves 5+ live clients from per-tenant JSON config. New clients are configured, not coded — days, not weeks.
5
Multi-supermarket data product (50K+ products)
We index 5 Paraguayan supermarket chains (Superseis, Stock, Areté, Casa Rica, Supermas) and maintain a price-comparison engine.
6
4-language production website
Full content in es / en / nl / pt, including nav, footer, legal pages, newsletter, contact forms. Mercosur + Europe + US coverage.
7
Self-hosted Rust + React video LMS
A built-from-scratch Coursera-clone: videos, quizzes, certificates, progress tracking. Rust + axum + React + PostgreSQL.
8
42 public repos as proof of work
Every claim is verifiable on GitHub. No other AI agency publishes this much of their actual work.
Recent activity
Public build log. Updated when something interesting goes live.
clientpreview
Dra. Gabriella dental site deployed in preview
Built apps/dra-gabriela/ in the monorepo with the content package from Ai-Whisperers/dentist. Public launch pending client data (phone, address, RUC, MSPBS).
researchopen-source
20-competitor market analysis published
Full analysis of the 20 AI agencies we benchmark against: services, pricing, threat level. Includes our 8 unique capabilities. Source: docs/research-sources.md (28 URLs).
salestools
28-capability sales sheet shipped
All 28 services with our pricing, market low/high, sources, and internal proof — as a Google Sheet via Apps Script or CSV import. Live in the company repo at Company/sales/.
infraaudit
VPS + Docker Swarm + Traefik fleet audit
52 services, 42 unique client sites, 55 Traefik hostnames, 8 overlay networks, 28+ volumes. Full breakdown in docs/infrastructure-audit.md.
The team
2 founder-led. 8 named AI bots. We work with a curated network for overflow.
Ivan Weiss van der Pol
Founder & CEO
Data Engineer, Automation Architect, QA Specialist. Founded AI Whisperers in November 2025. Mission: make AI practical for SMBs in Paraguay and LATAM. Built the @ai-whisperers/* npm package suite that powers 30+ client sites.
Next.jsTypeScriptPythonDocker SwarmTraefikMCP
💻 GitHubKyrian Weiss van der Pol
Co-Founder & Technical Director
n8n expert, AI integration specialist, technical architecture lead. Co-founder of AI Whisperers. Drives the systems side of the engineering studio — pipelines, integrations, automations, deployment pipelines.
n8nAI integrationAutomationTechnical architectureDeploy pipelinesWhatsApp automation
Get in touch
Free 30-minute strategy call. No sales pitch. No credit card.