Felipe Viana
Founder Engineer · Medellín, Colombia
WhatsApp +57 318 323 9505 · linkedin.com/in/felipevianas · github.com/vianasfelipe
Summary
Founder Engineer · Leading Health Tech teams to ship software that meets the clinical bar.
Two-time founder. I build at the seam of design, engineering, and the people doing the work. I work end-to-end — from pre-sales and solutioning to the first sketch and the system in production — and I help teams ship things that feel inevitable.
Experience
- EPAM
Leading Health Tech delivery for global clients.
I lead delivery on Health Tech engagements where the regulatory bar makes the craft sharper. I work end-to-end with clients: architecture, hands-on technical decisions, and the cadence that lets a team ship in a regulated environment.
- Medical Device Data System · Nov 2025 – Present. Delivery of a large-scale, cloud-native data platform that integrates diagnostic medical devices with an AI-assisted software ecosystem. The platform ingests biometric and imaging data (including DICOM), processes it through standardized pipelines, and powers downstream analytics, visualization, and decision-support features.Stack: Python, Databricks, +4 more
- AI agent in a regulated-adjacent domain
A personal project. I build an AI agent that reads unstructured records and drafts grounded outputs — using it as a testbed for tool use, evals, and everything I want to learn about agentic systems.
I picked a hard, structured problem on purpose: it forces every interesting agent question at once — how do you keep the LLM from making things up, how do you cite evidence, how do you build evals worth trusting, how do you design a rule engine the model can't override. Learn-by-doing, on the side, no commercial pressure.
- Multi-model agent design — different Claude models routed by task (extraction, comparison, rule synthesis).
- Defense-in-depth grounding: LLM output treated as untrusted candidate data, re-validated by deterministic tools.
- Evidence-linked outputs — every extracted item points back to its source text.
- Offline eval harness against expert-labelled cases, with recall and false-positive gates.
- End-to-end solo build: product surface, backend pipeline, rule engine, AWS deployment.
Stack: Anthropic SDK, FastAPI, PostgreSQL, AWS, Agent evals, +6 more
- AddSkin
A skin-tech product at the intersection of health and design.
Second venture. Closer to the Health Tech bar — clinical-grade clarity, consumer-grade craft.
- BookToFly
An online platform where travel agencies sold their services under their own brand and negotiations.
Co-founded and built end-to-end — product, engineering, integrations, and the first customer calls.
- Built the first version of the BookToFly API platform and pitched the product to early customers across Latin America.
- Led a team of 10+ people, sustaining a customer retention rate above 90%.
- Led the integration of major travel providers via web services — Amadeus, Sabre, Kiu, and Hotelbeds.
- Grew annual recurring revenue from $0 to $300,000 USD in the first five years.
- Expanded the customer base across Latin America and the United States.
Stack: MongoDB, Object-Oriented Programming, +3 more
Skills
- Design systems. Tokens, primitives, and documentation that survive contact with a growing team.
- Information design. Dense product UIs — clinical, financial, operational — that read at a glance, not a squint.
- Prototyping in code. I prototype in the real stack — Figma for layout, the codebase for behavior.
- Full-stack TypeScript. Next.js, tRPC/REST, Postgres, plus the AI/agent surface area (Anthropic SDK, tool use, evals).
- Systems that age well. Migrations as a first-class concern, contracts at boundaries, fewer abstractions.
- Performance & DX. Faster local builds, faster CI, faster onboarding — compounding wins for the team.
- Leading project teams. Currently leading Health Tech engagements at EPAM — multi-disciplinary squads, ambiguous problem spaces, regulatory constraints.
- Founding & hiring. Two companies, two founding teams. Calibrated rubrics, transparent leveling, frequent 1:1s.
- Written culture. Decisions, postmortems, and design docs that compound team knowledge.
Interests
Health Tech · AI agents · Typography & craft · Endurance sports