About
I get the full potential out of LLMs not by prompting harder, but by building the harness: gated codebases, physical guardrails, and agentic workflows that let AI agents ship production systems fast without breaking them. At WeAssist I own AI and data automation end to end, covering the ground of a small team, because diagnosing an inefficiency is only half the job if you cannot ship the fix yourself. Previously I built core systems for a multi-tenant ERP platform at Hatchit Solutions. BS Information Technology, Magna Cum Laude.
Work Experience
Skills
Ask an agent for something dangerous
A live model writes the change, constrained to a Zod schema. Deterministic gates then run on its output, in your browser, from src/lib/gates.ts. When a gate blocks, the reason goes back to the model and it rewrites. You are watching an agent get physically constrained, not persuaded.
Get in Touch
Reach me on LinkedIn or at markkennethbadilla@gmail.com. Happy to talk about agent harnesses, gated codebases, or anything you are trying to make impossible to get wrong.