Pedro Brun
Senior engineering partner
I build with founders, and help engineering teams put AI into production.
Senior software engineer based in Brazil, working at Savvy Wealth.
Engagements
Four shapes the work usually takes. Each is scoped before we start, with weekly delivery and a fixed end date.
- 01
AI in production
LLM features that ship and stay shipped. Retrieval, agents, evals, fallbacks, observability — the messy parts that decide if it survives past the demo.
- 02
Zero-to-MVP
For founders putting a first version of a product in front of users. Architecture, frontend, backend, deploy — owned end to end. AI built in where it earns its place.
- 03
Architecture review
One pair of senior eyes on the plan before it turns into an expensive rewrite. Scaling calls, design reviews, and a written report you can act on.
- 04
Team training
Hands-on training and pair-programming for engineering teams adopting LLMs, agents, and evals. Built on production-ready patterns, not toy examples.
How I work
Short, scoped, and written down. Most engagements run four to twelve weeks.
- 01/03
Scope call
30 minutes, no charge. Frame the problem, sanity-check the plan, see if we fit. You leave with a sharper next step either way.
- 02/03
Focused engagement
Fixed scope, weekly delivery, written updates every Friday. Code in your repo by week one. Async first, calls when they help.
- 03/03
Handoff
Documentation, walkthroughs, and a follow-up window so your team owns it after. No retainer needed unless you want one.
Selected work
Production systems across fintech, commerce, and consumer. Some named, some under NDA.
- 01
Voice-AI platform
Mapping and activating high-value connections across professional networks.
- 02
AI collections agent
WhatsApp Business product used by enterprises in Brazil to recover delinquent receivables at scale.
- 03
Reclaim
iOS app helping people break compulsive habits — content blocking, AI chat support, progress tracking. Co-founded and built end to end.
- 04
AI investment copilot
Reading regulatory filings with OCR and LLMs, surfacing what changed and what it means.
- 05
Kafka event backbone
For a live-commerce platform, adopted across four engineering teams as the system of record.
- 06
B2B procurement checkout
Handling single transactions up to R$5M with rebuilt reconciliation and fraud guardrails.
- 07
Product ingestion pipeline
Processing 500K+ requests per day from partner retailers, with backpressure and replay.
- 08
Sold-out NFT drop
Built and shipped for one of France’s largest newspapers; sold out on release.
Working on something? Let’s talk.
Putting AI into a product, scaling past a first version that works, or want senior eyes on a plan? Send a paragraph about the project — where it is and where it’s stuck.
↳ I read every message and reply within 48 hours.