Are you ready to join Connecticut Innovation’s vibrant community of innovators? Connecticut Innovations (“CI”) is Connecticut’s strategic venture capital arm, and we are passionate about serving our portfolio of 220+ companies across various industries, with strengths in life sciences, technology, and climate tech.
Come join Mavalis!
About Mavalis
Ultra-luxury travel membership connecting UHNW members with a curated network of world-class Expert Travel Partners (ETPs). Each member has a dedicated Mavalis Expert (ME). Seed funded, founding cohort onboarding, manual service already delivering trips competitors do not match. Now we digitalize it into an AI-enhanced platform, one workflow at a time. Stop planning. Start arriving.
Founding Engineer
You build the three product surfaces (Member Portal and App, ME Console, ETP Console) and the Intelligence Layer that connects them, on a modern web and mobile stack, deeply grounded in modern LLMs. You ship every week to real members and real ETPs. You set technical direction with the founders.
The 12-Month Mission
You are our hire number one on the engineering side.
- Phase 1 | Sep to Dec 2026: trust foundations, first AI workflow in production with the founding ME, secured-login Member Portal v0 live for founding members.
- Phase 2 | Jan to Mar 2027: ME Console v1, AI copilot grounded in the Mavalis graph.
- Phase 3 | Apr to May 2027: ETP Console v1 with inventory tools, booking workflows, commission visibility and Mavalis-provided payment rails.
- Phase 4 | H1 2027, target June: Member iOS App and Platform V1, three surfaces live on the Intelligence Layer.
Every phase starts with discovery alongside the MEs, the ETPs and a small group of founding members. You build what you learn.
What You Own
- Platform and Intelligence Layer | End-to-end architecture across three surfaces (Member Portal and App, ME Console, ETP Console) and the Intelligence Layer that connects them (Member Digital Twin, ETP Digital Twin, Trip Graph). You set direction and ship weekly. Accountability: the platform runs and scales from tens to thousands of members without a rebuild.
- AI Systems | Retrieval grounded in the graph, agents on structured member and ETP context, evaluation harness that turns every ME override into training signal, LLM observability from day one. Accountability: shipped workflows are measurably better than the analog alternative and the ME trusts them on real trips.
- Three Surfaces | Member Portal v0 then Member App (curation, not choice). ME Console (Member 360, briefing workflow, in-cockpit copilot). ETP Console (pre-qualified briefs, inventory, booking, commission, payment rails). Accountability: each surface ships to real users at its target phase, used on real trips with real money.
- Trust and Security Posture | Dedicated database project, row-level security, column-level encryption on sensitive fields, per-member envelope encryption with KMS-managed keys, audited decrypts, zero-retention DPAs with every AI vendor, GDPR native, SOC 2 Type I by V1. Accountability: no member data can be decrypted alongside anyone else's, ever, and every access is retrievable in minutes.
Current Tech Stack
We have thought hard about the shape of the stack. We have not written the first line of code. These are our current leans and the reasoning behind them. If you would do it differently after your first weeks in the codebase, make the case and we change our minds. The application deliberately asks you to pick one thing here you would do differently.
- Language and Web | TypeScript end to end. Next.js on Vercel for both the Member Portal and the two internal cockpits, so the cockpits are real product surfaces from day one (not throwaway internal tools) and the eventual customer web experience is an evolution of the same codebase, not a rewrite. Monorepo, shared types, shared UI primitives.
- Mobile | Expo and React Native, iOS first for the Member App at V1.
- Database and Data | Postgres as the single source of truth across all three surfaces, hosted on Supabase. ORM (Prisma or Drizzle, engineer's call). Vector search for retrieval grounded in the Member and ETP Digital Twins, on Postgres extensions or a hosted vector service, engineer's call after the discovery phase.
- API and Service Layer | Some structured layer above the database for business logic (matching, retrieval scoring, audit trails). tRPC, a small Node service, Next.js API routes: engineer's call. What we care about is that there is a real, testable layer where business logic lives, not scattered auto-generated CRUD.
- Auth. | Clerk with magic-link plus passkey. Enterprise-grade session management from day one because our members are not tolerant of friction and our compliance posture cannot tolerate anything less.
- AI (what you build with it) | AI-powered features across the three surfaces: the ME copilot inside the ME Console (member 360, briefing extraction, preference deltas, draft follow-ups); the ETP-facing brief matching and ranking that pipes qualified opportunities into the ETP Console; the matchmaking layer between members, trips and ETPs grounded in the Intelligence Layer; retrieval and reasoning over the Member and ETP Digital Twins; the evaluation harness that turns every ME override into training signal.
- AI (the model stack) | Anthropic and OpenAI under zero-retention terms as the default LLM providers in Phase 1. Architecture is model-agnostic from day one, and from Phase 2 we introduce smaller specialised models (open-source or open-weights, self-hosted) where cost, latency or data posture outweighs the frontier quality gap. From Phase 3, classical ranking and graph-based models trained on the Trip Graph itself, complementing the LLM stack. Transcription: Deepgram or comparable. Observability: Langfuse or comparable. Structured outputs, evaluation harness, cost and latency instrumentation from the first workflow.
- Infrastructure and Security | AWS KMS for per-member envelope encryption. Cloudflare in front. GitHub Actions for CI, staging environment mandatory, feature flags for every release. Payments: Stripe with the two-sided flow. Email: Resend.
Who You Are
- Behaviors | Owner mindset. Product taste. Bias to ship. Quiet confidence. Discretion. Intellectual curiosity. Comfortable being first, with the ownership, freedom and accountability that entails.
- Experience | 6+ years engineering, 2+ years shipping LLM-based products in production. Track record of technical ownership at a startup, or as a founding or early engineer where the platform grew under your hands. Deep hands-on with modern LLMs: retrieval, agents, evaluation, structured outputs, cost and latency at scale, observability. Strong full-stack on TypeScript, Next.js, React (Native welcome). Comfortable owning a Postgres schema and the API layer above it. Solid data engineering: schema design, event modelling, relational and vector data patterns. Comfortable with messy third-party integrations (booking APIs, payments, KYC, travel supplier feeds). Strong English.
Bonus: background in travel, hospitality, luxury or any high-touch consumer context.
Mavalis is committed to employment equity and building a diverse workforce. We welcome and encourage Indigenous applicants, people of color, all genders, 2SLGBTQ+ and persons with disabilities to apply. Accommodations are available on request for candidates taking part in all aspects of the selection process.