Kategorien durchsuchen
Entdecken
Fiverr Pro
Deutsch
$
USD
I will build a custom ai agent for finance accounting and audit
Sun X
Über diesen Service
Most prompt engineers don't know finance. Most accountants don't know AI. I do both.
I'm Kenny a CPA and Bar-qualified CFO at a state-owned financial group. I also build production LLM agents via Claude Code as my actual work, not a side hobby.
Why this matters: generic LLMs hallucinate badly in finance. They confuse IFRS 9 with IFRS 15, miss related-party flags, mis-classify leases, and confidently produce numbers that don't reconcile. A prompt engineer without domain depth can't catch any of this.
What I've built: CPA-Agent autonomous audit & equity research agent on OpenClaw, with layered prompts that cut hallucinations on gray-area accounting Domain-specific eval sets benchmarked against expert CPA judgment GlobalFinCloud macro dashboard via v0.dev + Claude Code
I help you:
- Build custom finance/audit AI agents
- Engineer layered system prompts (IFRS / GAAP / PRC)
- Design domain evaluation sets
- Audit existing pipelines for failure modes
Good fit for: AI startups building vertical finance agents, accounting firms experimenting with LLMs, founders who hired a prompt engineer and got outputs that "sound smart but are wrong."
- Message me before ordering .
Lerne Sun X kennen
Sun X
Chief Financial Officer
- AusChina
- Mitglied seitMärz 2026
- ⌀ Antwortzeit13 Stunden
Sprachen
Englisch
CPA, Bar-qualified CFO & ACCA Strategic Professional candidate with 10+ years of expertise. I bridge the gap between rigorous IFRS compliance and AI automation.
As a Financial Engineer, I develop custom Python-based AI tools for SEC filing parsers, automated equity research, and algorithmic fund management. My edge lies in transforming complex data into actionable insights through technology.
Need IFRS reporting, AI-driven investment workflows, or financial modeling? I deliver solutions that are legally rigorous, technically sound, and technologically advanced.
Mein Portfolio
FAQ
Which LLMs do you work with?
Primarily Claude (Sonnet/Opus) via Claude Code, plus GPT-4 family and Gemini. My architecture is model-agnostic — the layered prompt structure transfers across providers. Tell me your stack and I'll work in it.
I'm a non-technical founder. Can you actually deliver working code?
Yes. I build via Claude Code and v0.dev — meaning I architect with prompts and ship working pipelines, not just slide decks. You'll receive runnable code (Python or JS, depending on your stack) plus a clear handoff doc your engineers can extend.
How is your prompt work different from a generic prompt engineer?
A generic prompt engineer optimizes for output that looks right. I optimize for output that is right under CPA scrutiny. The difference shows up when your agent has to handle gray-area judgment — revenue cut-off, consolidation scope, fair-value measurement — where a confident wrong answer is worse t
Will my data and prompts stay confidential?
Yes. I sign NDAs by default for any project. Your prompts, eval sets, and pipeline architecture stay yours.
Can you handle non-English finance work?
Yes — I work fluently across English, Mandarin, and traditional Chinese. PRC GAAP, CSRC disclosure rules, and HK Listing Rules are part of my actual job, not just a research interest.
