Anuj Sadani

People come first because they are the only part that compounds. Systems decay, frameworks get replaced, and the roadmap you defend today will be rewritten. What remains is the judgement in the room — and judgement is grown, not hired in finished form.

So I treat mentoring as the actual work, not the overhead around it. If the team ends a year abler than it started, the systems tend to follow.

In practice mentoring 12+ Python and ML engineers, hiring for density over headcount.

Business sits in the middle because it is the honest translator. It is where an idea has to survive a question engineers would rather skip: what does this change for someone who does not care how it works?

Engineering that cannot answer that gets decided for it, by people further from the code. I would rather make the argument myself, with a number attached.

In practice turning requirements into technical OKRs, and hype into bets with a cost attached.

Technology comes last, and I say that as the person who enjoys it most. It is the layer with the best feedback loops and the least ambiguity, which is exactly why it is so tempting to start there — picking the stack is easier than admitting you have not agreed on the problem.

Put last, it becomes leverage. Put first, it becomes a very expensive opinion.

In practice enterprise RAG, agent tooling on MCP, fine-tuned open models, model routing, evals, and the org-wide Python stack.