Senior Full-Stack Engineer ยท Mexico City
I'm David Behar. I build the platform that helps dozens of developers create React apps, 15+ microfrontends, and 75+ serverless functions. 10+ years of shipping TypeScript at scale, and I still get excited about a clean deploy pipeline. Lately I'm building AI agents because one career at a time felt too easy.
With my team, I build the platform that other engineering teams ship on top of. React microfrontends, serverless backends, shared auth SDKs, telemetry packages. When something I ship breaks, 70+ engineers across 3 time zones let me know in English, Spanish, and a wall of Slack pings.
I broke a ~4MB monolithic bundle into 20+ micro-frontends using Webpack Module Federation, which cut initial page loads by ~35%. I own 75+ Lambda functions across 15+ microservice backends. I built 5+ internal libraries that brought developer onboarding down from roughly 2 months to about 2 weeks.
Before Autodesk, I ran an agency called Qubit Works, growing the team from 2 to 10+ engineers and delivering 20+ client projects. That's where I learned the hardest part of engineering: turning "I want something like Uber but for pharmacies" into an actual spec.
TypeScript
React
Redux
Next.js
Webpack
Storybook
Tailwind CSS
Node.js
AWS Lambda
GraphQL
Jest
GitSplit a ~4MB monolith into 20+ micro-frontends with Webpack Module Federation. Initial page loads dropped ~35%. Friday deploys stopped being a source of dread.
Incremental builds, parallelized pipelines, independent MFE releases. Went from 12+ deploys per shared component change down to ~5. The team actually started shipping on Fridays again.
Built 3 large-scale React apps (water telemetry, CCTV monitoring, asset management) with 15+ microfrontends and 7 shared libraries. 70+ engineers across time zones depend on these daily.
Upload/download workflows, compliance validation, event-driven pipelines. All on AWS Lambda, API Gateway, and EventBridge across 15+ microservice backends. Built an auth SDK that cut integration time in half.
Multi-agent systems for code generation, automated review, and test scaffolding. I use Claude Code every day and I'm wiring up autonomous pipelines that handle the repetitive work so engineers can focus on the problems worth solving.
Conversational agents that qualify leads and handle follow-up for B2B companies over WhatsApp. In Latin America, WhatsApp is the enterprise communication channel, so that's where the bots live.
Taking messy inputs (documentation, Slack threads, meeting recordings) and structuring them into searchable, queryable knowledge bases. Tribal knowledge is a liability when it only lives in someone's head.
Check out my blog for thoughts on engineering, architecture, and AI.
Browse all posts