Srikanth Bandaru

Work

• FarmersAdding a country without adding a codebaseI lead the squads behind our farmer sites in North America, Australia and New Zealand. For New Zealand, the business expected an expensive custom build. I pitched reusing Australia’s build instead, as a shared platform where each country is a configuration. I wrote its core, and my squad launched New Zealand on it in 3 weeks. That saved about $200K and won a company recognition award. Ruveon runs on it too, and Canada, Bangladesh and the US are next.
• SalesFrom hackathon to 1,200 sales reps in 64 daysI’m the global experience lead for our sales apps in seven countries, and every team on them ships on a platform I built. In the US it went from a five-day hackathon build to 1,200 reps using it in 64 days, with an AI visit planner I co-built.
• Design systemA design system that doesn’t need its own teamMost design systems need a team to keep them alive. I proposed ours and built it on shadcn. Early on, I reversed my own decision to own the core components. Brand styling now lives in wrappers, so upstream fixes flow in untouched. One token pipeline keeps a brand the same on web and mobile. Ideation to production took under a month. It’s now the official design system for our Customer Experience Platform, and more than 15 teams build on it.
• Agentic SDLCBuilding an agentic SDLC that agents can’t gameAgents wrote plausible code that ignored the decisions we had made. I built Build OS to give them our decisions, repos and plans, with CI that validates that context. When agents gamed its coverage score to a perfect 1.00, I replaced the score with an adversarial audit. Now I’m turning it into a plugin engine, so any team on our Customer Experience Platform can run an agentic SDLC on its own tools. Agents take work a person marked AI-ready all the way to a pull request.
• AI toolingDesign-system tools for coding agentsI built an MCP server that helps coding agents use Element, a Bayer design system used by more than 100 teams. The Element team adopted it as their official server. I also built Ralph, a CLI that screenshots an AI-built component, pixel-diffs it against its Figma design and audits its HTML against design-system rules. It then has the agent fix the top issues and retry until the check passes or stops improving.
• CommunityBeyond my own teamsI’ve taken our AI work beyond my teams: workshops, a talk at Bayer’s ignITe conference, and demos to our principal engineers, Crop Science IT’s leadership and our CEO. Outside Bayer, I work with Vercel on how Bayer runs on their platform, and Datadog merged a fix to their browser SDK from my repro.

Full résumé