Solutions / VP Engineering
Every integration request is time not spent on product. Tray.ai takes the backlog off engineering without trading reliability for speed.
Every integration request is time off the product roadmap. HackerOne quadrupled delivery speed — 2–3 months became 2–3 weeks. Airbnb compressed an 8-week MuleSoft project to one week. When leadership asks for AI agents in production, it’s not a new vendor evaluation — the same platform already running the integrations handles that too.
700+ connectors handle the SaaS surface. Engineering stops writing bespoke connectors for every ops request.
JavaScript and Python steps drop in anywhere — no separate build environment, no context switching.
Merlin Agent Builder and Agent Gateway live on the same platform as the integrations. No separate AI vendor.
SOC 2 Type II, HIPAA, GDPR, RBAC, full audit trails per run. The platform's posture covers security review.
Teams the vp engineering role tracks for fit, scale, and proof.
Tech / SaaS
GitHub's marketing automation platform and CRM didn't speak fluently to each other, so leads from a 36M-developer community were arriving wi …
Read the story →Tech / SaaS
HackerOne's customers respond to security vulnerabilities across dozens of different tools, so integrations are core to the product — but ea …
Read the story →The use cases this role pursues first on Tray.ai.
Integration without the integration backlog.
Replace MuleSoft, Boomi, or whatever stitched-together iPaaS is slowing you down. Tray.ai connects 700+ apps and prepares data in flight — at enterprise scale, in days not quarters.
Manual work, automated.
Automate the manual work IT hasn't gotten to — across 700+ connectors, with audit trails baked in. Without waiting for engineering.
Ship production agents, not pilots.
Most AI projects stall in pilot. Tray.ai is the path to production agents — reasoning, acting, and audited across your real systems.
More integrations. Less engineering.
Ship integrations with the partners that matter to your customers — at the cadence customers actually expect. Bynder, Cvent, Cue, HackerOne and Podium replaced a years-long engineering backlog with a marketplace where new partner integrations land in days, not quarters.
The build case for custom integrations typically wins on fit and loses on time and maintenance. In practice — HackerOne went from 2–3 month custom builds to 2–3 week configurations. Airbnb compressed an 8-week MuleSoft project to 1 week. The question to model is total cost over 3 years — build cost, maintenance cost, incident cost, opportunity cost of engineers on integration work.
Full. JavaScript and Python steps are first-class — engineers can drop into code anywhere in a workflow. The platform handles the connector surface, retry logic, error handling, and execution infrastructure. Engineering focuses on the logic that's specific to the business, not the plumbing that isn't.
Agent Gateway gives engineering managed MCP servers with RBAC, rate limits, observability, and audit logs. The same governance model that covers the integration layer extends to AI agents — no separate security review, no separate credential management. J.W. Pepper's Marcus Dubreuil described the result as workflows becoming microservices — composable, versioned, and auditable.
The pattern is phased, not rip-and-replace. Start with the highest-friction integration (the one with the most support tickets or the most manual intervention) and prove the platform in production. Airbnb, Yext, and Crowdstrike all followed this pattern before broader rollout. The rollout accelerates as the team builds confidence in the platform.
Walk through your real systems with someone who's been there.
