Tong L.

Key Skills

Software

No software listed

Top Subject Matter

Financial audit

Regulatory Compliance & Risk Analysis

business operations

Top Data Types

Text

Document

Top Task Types

Prompt + Response Writing (SFT)

RLHF

Evaluation/Rating

Question Answering

Freelancer Overview

As a former Big 4 (Deloitte) IT Audit and Forensic Data Analytics Director holding CISA and CGA credentials, I bring a zero-tolerance approach to logical inconsistencies and factual inaccuracies. My extensive background in IPO prospectus verification (circle-up), complex SOC compliance, and big data anti-fraud investigations translates perfectly into high-tier AI data labeling and LLM output evaluation. I specialize in rigorous fact-checking, cross-verifying complex financial and IT system logic, and identifying subtle data anomalies. This high-stakes auditing mindset is essential for detecting AI hallucinations, auditing data assumptions, and ensuring the absolute accuracy and reliability of AI-generated responses. Building on my domain expertise, my recent leadership role at an AI startup directly involved AI alignment, SFT (Supervised Fine-Tuning) dataset creation, and RLHF (Reinforcement Learning from Human Feedback) for financial and audit AI Agents. I have proven experience in translating complex, implicit business rules into structured ontologies and designing high-quality Chain of Thought (CoT) prompts to enhance LLM reasoning capabilities. Furthermore, my hands-on work in evaluating agentic workflows, conducting red-teaming for boundary edge cases, and fine-tuning AI behaviors ensures I am fully equipped to handle advanced, logic-dense annotation tasks that go far beyond basic data labeling.

Labeling Experience

AI Agent Trainer & Evaluation Specialist

TextRLHF

Participated in the training loop for financial and audit-focused AI agents through expert-level human feedback. Evaluated the accuracy, consistency, and compliance of agent outputs in complex computational and decision-making scenarios. Identified and corrected model hallucinations and logical inconsistencies in high-stakes business tasks. • Evaluated multi-step agent workflows for correctness and tool usage in automated tasks. • Audited agent decision-making for IT controls and regulatory compliance. • Provided detailed human feedback to refine model behavior and output quality. • Contributed to model alignment by diagnosing and solving edge cases and adversarial challenges.

Present

Director of Domain Knowledge & AI Agent Training (AI Training & SFT Labeling)

TextPrompt + Response Writing (SFT)

Led the development of high-quality instruction fine-tuning (SFT) datasets for aligning large language models with complex business logic. Designed robust guidelines and rubrics to ensure all labeled data adhered to strict accuracy and clarity standards. Conducted ontology and taxonomy engineering to structure domain knowledge into explicit, model-understandable rules and workflows. • Structured human tacit knowledge in finance, compliance, and operations into explicit rules for LLM training. • Developed multi-step Chain of Thought (CoT) prompts for advanced professional reasoning. • Ensured 100% accuracy by implementing stringent data labeling protocols. • Oversaw a team in producing gold-standard ground truth datasets for SFT and domain alignment.

Present

Education

Work History