Igor Eduardo

Senior AI Product Engineer

Austin, TX

About

I build production AI systems for healthcare. I solo-engineered a clinical AI platform from zero to production — 10 integrated modules, hybrid RAG pipeline over thousands of curated medical papers, self-hosted medical LLM, and healthcare regulatory compliance (HIPAA/LGPD-grade). Currently serving physicians in clinical pilot.

Before engineering, I spent 7+ years in M&A and venture capital, including a successful founder exit (acquired by the largest PR conglomerate in Latin America) and portfolio value creation for a R$1B+ AUM fund. This combination of hands-on engineering and business thinking is how I build AI systems that are technically rigorous, economically viable, and regulation-ready.

What I Built

DocMinds.ai — Clinical AI Platform
Founder & Lead Engineer | 2025 – Present

Production clinical AI platform for physicians. The doctor asks a clinical question in natural language and receives an evidence-based response with verified citations in under 2 seconds. Accessible via web and WhatsApp.

What I engineered: Hybrid RAG (BM25 + semantic + Reciprocal Rank Fusion) with zero citation hallucinations. Self-hosted medical LLM (AWQ quantized) with 4-model fallback chain achieving 80–95% cost reduction. 5-layer safety system with PII scrubbing, drug interaction checks, and regulatory disclaimers. 10 modules: clinical chat, lab interpretation (22+ analytes, deterministic classification), ECG analysis (ResNet-1D trained on 345K Brazilian ECGs), ambient clinical scribe (Whisper + speaker diarization), adaptive quiz engine (1,053+ real exam questions), spaced repetition flashcards (ts-fsrs), interactive clinical cases, scientific library, WhatsApp channel, and gamification.

0%Citation hallucinations
10/10Audit score
145+Automated tests
80–95%Cost reduction
<2sResponse time
94.9%Gross margin
minds.ai — Sovereign Medical AI Infrastructure
Founder & Lead Engineer | 2025 – Present

The AI infrastructure layer behind DocMinds. Proprietary model built on MedGemma, PaliGemma, and other open-weight foundations, fine-tuned with real Brazilian clinical interaction data. Five-layer safety agent, structured clinical knowledge fabric (national protocols, drug formularies, diagnostic codes, epidemiology), and data sovereignty — all processing on Brazilian soil.

Self-hosted LLM RAG Pipeline Healthcare Compliance Data Sovereignty

Previous Experience

Strategy & Operations — AVIV / OpenVC 2018 – 2024

Portfolio growth and value creation for R$1B+ AUM fund platform. Sourcing, due diligence, governance, crisis PMO. Notable outcomes: Neoway exit ~$360M (B3), Grão Direto seed-to-Series B ($0.46M to $15M).

Venture Capital M&A Portfolio Ops
Founder & CEO (Exited) — Kubrick.social 2019 – 2020

Built white-label orchestration platform automating campaign workflows. Clients included Uber, Rappi, Airbnb, Spotify via agency partners. Successfully exited to the largest PR conglomerate in Latin America.

Founder Exit SaaS Automation

Technical Skills

AI / ML
RAG (BM25 + Semantic + RRF) LLM Inference (Self-hosted) pgvector / HNSW Citation Verification Embedding Pipelines QLoRA Fine-tuning Model Routing Prompt Engineering Clinical NLP
Engineering
TypeScript (strict) Next.js PostgreSQL Python Node.js Supabase Redis REST APIs SSE Streaming
Infrastructure & Quality
Vercel / Railway / RunPod GitHub Actions CI/CD OpenTelemetry Langfuse (LLM Observability) Sentry Vitest / Playwright Semgrep (Healthcare SAST)
Healthcare & Compliance
HIPAA / LGPD PII Scrubbing (Presidio) CFM 2.454/2026 Drug Interaction Checks GRADE Evidence Framework FHIR Awareness
Business
Unit Economics M&A / Due Diligence GTM Execution Financial Modeling Stripe / Payments

Open Source

hybrid-rag-pipeline TypeScript / MIT

Production-grade RAG combining BM25 + semantic search + Reciprocal Rank Fusion with authority boosting and recency weighting.

citation-guard TypeScript / MIT

Post-generation citation verification for RAG systems. Eliminates hallucinated references before they reach users.

model-router-chain TypeScript / MIT

Smart LLM routing with automatic fallback, complexity scoring, and cost optimization. 80–95% cost reduction vs single-model.

Education

Y Combinator — Startup School
Stanford University — Executive Program, Design Thinking (2019)
UC San Diego — International Program (2024)
International American University — Business Administration (in progress)

Languages

English (Fluent) • Portuguese (Native) • Spanish (Fluent)

Connect