I am a nontraditional candidate: seven years owning a real service business, a decade-plus in field technical work, and a recent AI build sprint that produced live products, agent systems, search/indexing infrastructure, and thousands of commits. My angle is not pretending to be a conventional ML researcher. My angle is execution, people judgment, and making complex AI work understandable enough for teams to act on.
Research Manager may be a stretch on paper. The real fit is AI specialist, AI engineer, app builder, research operations, interpretability tooling, technical program leadership, or an adjacent builder role where a team needs someone who can keep ambiguity moving.
I do not have the standard academic ML path. I do have a visible record of shipping, learning in public, and building systems that force me to ask the same kind of questions interpretability cares about: what is happening inside the system, what signal matters, and how do we make the behavior inspectable enough to trust?
For modern AI specialist, AI engineer, app hiring, and tech hiring roles, years in an older tech lane are not enough by themselves. The useful signal is whether someone can direct AI systems, verify outputs, ship working tools, and explain the workflow clearly to other people.
A person can have a long technical background and still not know how to use agents, structure prompts, verify source chains, manage model failure, deploy quickly, or turn AI output into a reliable product. My proof is current: repos, commits, deployed sites, traffic, and working workflows. I am not selling a title. I am showing the work.
The point is not that every number is perfect. The point is that there is enough public, checkable surface area to see the work is real.
These are the projects I would want a recruiter or research lead to click first.
Public crash and traffic data, evidence packets, and a real-scale browser simulation for exploring intersection risk. The useful part is the operating pattern: gather messy public data, make it inspectable, and present it with guardrails.
Quotes, workflow, operational tooling, Cloudflare deployment, domain binding, and real small-business use cases shaped from lived operator pain.
Agent workflows, local/cloud AI, project memory, dashboards, source verification, and a self-learning article/workflow fleet. The repo history shows the learning curve and iteration speed.
Search and investigation interface oriented around source quality, public records, and cited AI briefs. Built to make claims inspectable instead of just persuasive.
Working database, auth, dispatch/payment-confirmation workflows, and patent-pending product thinking. Kept private, but part of the broader build record.
I am useful where the system is too messy to read at first: many signals, unclear ownership, high stakes, and a team that needs the next concrete step.
My day-to-day stack crosses local and hosted models - Ollama, Qwen, Claude, ChatGPT, Kimi, Minimax, Grok, and other model/API surfaces - so the operating system matters as much as the model choice.
I believe useful AI growth requires security hygiene, auditability, clear source attribution, human accountability, and respect for established AI safety and governance principles from Berkeley-affiliated AI safety work and other credible research, standards, and policy organizations. The practical rule is simple: move fast, but do not hide uncertainty, skip security, or pretend generated output is verified evidence.
I own and operate Austin's Affordable Garage Doors in Fremont / the East Bay. That matters for hiring because it adds customer pressure, trust, sales, scheduling, field diagnostics, reputation, and accountability to the technical story.
AI hiring needs people who can ship and explain under pressure. Running a local service business for seven years trained the people side: plain-language explanations, urgent triage, trust repair, customer follow-through, and practical judgment. The AI portfolio shows those same operator habits applied to software, agents, search, and deployment.
Research operations, interpretability tooling, AI product operations, agentic workflow builder, technical program execution, AI systems prototyping, and founder-operator roles where fast learning and shipped proof matter.
The strongest proof is public and inspectable: github.com/pain2hustle/APIARY, github.com/pain2hustle/service-pricer, apiarybee.com, servicepricer.app, Cloudflare traffic, and the resume PDF linked on this page.
No. The honest claim is AI-native founder-operator and systems builder. The best-fit lane is adjacent to research teams: making workflows visible, building tooling, asking practical questions, and moving ambiguous work into usable systems.
Seven years running a real garage-door business created pressure-tested customer judgment, communication, sales, hiring, scheduling, and accountability. The AI work adds current technical execution on top of that operator base.
This page is meant to support the resume: the resume gives the formal version, the links prove the build history, and this page explains the nontraditional fit without hiding the nontraditional part.
Best target framing: AI specialist, AI engineer, AI team hire, app builder, or AI-native founder-operator for research operations, interpretability tooling, technical program execution, or a high-ambiguity builder role adjacent to a research team.