Skip to content
Vivek Tumrukota

What I’m building now

AI tools are table stakes. AI-native orgs are the edge.

Most companies have rolled out AI tools. Far fewer have changed how software actually gets built. That second part is my job.

Org-wide · 650+ person BU

The Software Factory

I’m driving Superhuman’s Software Factory effort globally: an agentic spec → plan → build → eval → review → ship pipeline that multiplies the true throughput of the engineering org. We measure shipped outcomes and cycle time, not lines of code or tokens burned.

  • Agent-ready repos, specs and CI
  • Human review where judgment matters
  • Throughput measured end to end

Superhuman Mail · 70+ engineers

Mail AI Guild

I lead the champions network that turns individual AI wins into team defaults, through shared patterns, office hours and weekly show-and-tells.

Beyond engineering

Everyone ships

Product, design, support and ops contribute production code safely, with guardrails, templates and review paths built for non-engineers.

Leverage

Skills & patterns library

Reusable agent skills, prompts and workflows that make the best way to do something the default way.

Quality

Evals, not vibes

Golden datasets, LLM-as-judge rubrics and online metrics for AI features and agent-written code.

Strategy

Frontier vendors

Model and tooling bake-offs, and contracts negotiated with frontier AI vendors to tie spend to leverage.

Staying ahead

Field intel

Models, tools and trends tracked weekly, then brought back to the org as patterns people can use on Monday.

  • Code with Claude
  • Cursor’s conference
  • Weekly model & tool radar
  • SF AI community
  • Hands-on evals of new models