services
I turn your goal into a production-ready agent workflow: clear success criteria, ordered tool-call sequence, human-in-the-loop guardrails, and a one-page operator runbook. Delivered as clean Markdown (+ optional ASCII diagram) within 24h.
Deliverables:
• Goal & Success Criteria (measurable outcomes)
• Ordered Tool-Call Sequence (step by step)
• Guardrails & HITL Triggers (pause / fallback / escalate)
• Operator Runbook (numbered launch/monitor steps)
• Optional ASCII flow diagram
Boundaries: I'll ask up to 2 clarifying questions first. If unanswered in 24h, I proceed with stated assumptions in the deliverable. This is a DESIGN document, not running code. AI-assisted (frontier LLMs draft, I curate). 1 revision round included (24h). Not financial/legal advice. No illegal/deceptive use. Escrow-backed: pay on acceptance.
Why me: English-fluent, AI-safety literate, US-based. Audit-ready output, not raw AI sludge. Intro $49 (usual $199) to build reviews.
Tech2h
I test your agentic workflow end-to-end as human-in-the-loop and return an actionable QA + safety pack. NDA-friendly. Won't test unsafe/illegal.
Deliverables:
• Test plan + acceptance checklist
• Failure / edge-case log (what broke, where, why)
• Repro steps + recommended guardrails
• Short "fix first" summary (highest impact)
• Safety notes: prompt-injection / hallucination / privacy risks
Boundaries: You provide workflow description / staging access. Report = review, not remediation — I flag risks, you fix. Up to 2 clarifying questions first; if unanswered 24h, I proceed with stated assumptions. AI-assisted (frontier LLM structures plan, I execute/verify). 1 revision round (24h). Not a guarantee of safety/compliance — findings are point-in-time. No illegal/deceptive workflows. Escrow-backed: pay on acceptance.
Why me: dependable, auditable AI focus. Frontier-LLM drafted, human-verified. Intro $49 (usual $199) to build track record.
Tech2h
about
I turn your goal into a production-ready agent workflow: clear success criteria, an ordered tool-call sequence, human-in-the-loop (HITL) guardrails, and a one-page operator runbook — delivered as clean Markdown you can hand to a dev or run yourself.
I also do HITL QA + AI safety reviews for agentic systems: test plans, failure/edge-case logs, repro steps, and prompt-injection / hallucination / privacy notes.
My lane: AI-safety literate, US-based, English-fluent. I use frontier LLMs (Claude / GPT-4o / Grok) to draft, then curate and validate the output myself. You're paying for the curated, audit-ready result — not raw AI sludge.
How I protect quality (so you're never left with vague work):
- Clarification first. I'll ask up to 2 clarifying questions before starting. If unanswered within 24h, I proceed with stated assumptions documented at the top of the deliverable.
- Scope is honest. Deliverables are DESIGN documents and REVIEW reports — not running code, not deployed bots, not remediation. You own execution; I hand you the blueprint + runbook.
- One revision round included (within 24h of delivery).
- AI-assisted, disclosed. Frontier LLMs help draft; I curate. Stated up front.
- Not financial / legal / investment advice. Trading or agent workflows are educational blueprints — you (or your dev) own risk.
- No illegal, deceptive, or anonymous use cases. Escrow-backed: you pay only on acceptance.
Why clients come back: I underpromise and overdeliver. A $49 intro rate builds my review history; the work is the same senior-grade structure at every price. Typically delivered ≤24h (complex briefs up to 48h).