Hayder Zaidi

AI Trust & Safety, Compliance, and Model Evaluation

I work where AI safety policy meets production reality: model outputs, reviewer judgment, escalation pressure, and sensitive-content risk. My job has usually been to find the part of the system that is quietly drifting and make it measurable enough to fix.

Professional Profile

AI Trust & Safety and compliance professional with 6+ years across AI model evaluation, red-team prompt review, child safety, fraud, e-commerce risk, and large-scale moderation operations for TikTok, Meta programs, Highspring, and Handshake AI.

I have worked the full chain: frontline review, QA calibration, SME escalation, team management, vendor quality, LLM prompt evaluation, and policy alignment. That matters because the real problems rarely sit in one box. A model miss can be a policy problem. A policy miss can become a training problem. A training problem can become a queue-health problem by Friday afternoon.

My work is strongest in ambiguous, high-sensitivity environments: child safety, self-harm, sextortion, spam/scam, fraud, and AI outputs that need careful review rather than a fast label.

Languages: English, Arabic

AI / Moderation

  • AI Red-Teaming
  • LLM QA Test Design
  • Model-Policy Alignment
  • Prompt & Response Evaluation
  • Content Moderation Systems

Trust & Safety

  • Child Safety Review
  • Sensitive Content Escalation
  • Policy Governance
  • High-Risk Queue Management
  • Enforcement Consistency

Leadership

  • Team Management
  • Mentorship
  • Vendor Calibration
  • Knowledge Base Design
  • Cross-Functional Collaboration
  • Executive Reporting

Technical / Analytics

  • SQL
  • Google Data Analytics
  • Cybersecurity Fundamentals
  • Root-Cause Analysis
  • Trend & Variance Analysis
  • Dashboard / Metrics Review

Professional Experience

Handshake AI – Remote

May 2026 – Present

AI Content & Safety Evaluator – Child Safety & Sensitive Content

Evaluate high-sensitivity AI prompts and model responses where policy, legal escalation, and human judgment all have to line up. The work is deliberately careful: the goal is not fast labeling, but consistent safety judgment under pressure.

  • Review AI-generated prompts and responses involving child safety, CSAM, self-harm/suicide risk, and other sensitive-content categories
  • Assess prompts and outputs for policy violations, safety gaps, and compliance risk across Trust & Safety evaluation workflows
  • Escalate high-risk cases involving potential NCMEC reporting and law-enforcement referral paths according to legal and regulatory requirements
  • Collaborate with prompt engineering and policy teams to improve AI safety guidelines, evaluation frameworks, and enforcement consistency
  • Produce safety findings that highlight model vulnerabilities, recurring policy gaps, and emerging risk patterns

Highspring – Remote

September 2024 – Present

Senior AI Guidelines & Compliance Evaluator

Support AI safety and compliance evaluation for enterprise AI systems, with a focus on prompt red-teaming, model-output review, evaluator calibration, and root-cause analysis when guidance is being applied inconsistently.

  • Evaluate AI-generated prompts and model responses against safety and compliance policies in ambiguous content scenarios
  • Conduct red-team exercises to surface harmful outputs, policy violations, and prompt patterns that expose model weaknesses
  • Audit vendor evaluation quality, policy adherence, and calibration performance across evaluator groups
  • Lead coaching and training for underperforming evaluators to improve policy accuracy and decision consistency
  • Document findings and produce compliance reports that help strengthen AI safety evaluation processes

TikTok – Austin, TX

January 2024 – March 2026

Sr. Quality Assurance Analyst – E-Commerce & LLM Programs

Worked at the intersection of AI moderation, seller compliance, QA systems, vendor performance, and policy enforcement in a high-volume social commerce environment.

  • Redesigned QA workflows and escalation paths, improving operational efficiency by 9% and reducing late-stage reversals
  • Partnered with AI and product teams on LLM QA testing, improving model-policy alignment accuracy by 12%
  • Delivered structured training and policy guidance to 200+ sellers, contributing to a 25% reduction in violation rates
  • Maintained 99% QA pass rate during rapid policy updates and enforcement changes
  • Led vendor-management review rhythms across moderation teams in Malaysia and Mexico, bridging vendor moderators, prompt engineers, and developers
  • Served as point of contact for child safety, spam/scam, sextortion, and e-commerce policy workflows, including high-risk escalation paths
  • Collaborated with Governance and Compliance teams on EU Digital Services Act (DSA) readiness and enforcement alignment

Accenture (Meta) – Austin, TX

July 2022 – December 2023

Trust & Safety Team Manager

Led high-risk moderation operations supporting Meta platforms, balancing speed, accuracy, and policy consistency across sensitive content categories.

  • Managed a team of 15–20 reviewers across child safety, sextortion, spam/scam, payment fraud, and other high-risk queues handling 8K+ daily tickets
  • Designed decision trees and escalation frameworks that reduced escalation volume by 18% and improved first-pass accuracy to 99%
  • Scaled Arabic-language anti-terrorism moderation operations from 6 to 40 reviewers, contributing to contract renewal
  • Implemented QA calibration, coaching loops, and knowledge systems reducing error rates by 40% and handle time by 20%
  • Handled 20K+ high-risk reports with less than 2% escalation, maintaining platform safety and compliance standards
  • Escalated high-risk child safety cases to NCMEC and law-enforcement pathways according to legal and regulatory requirements

Accenture (Meta) – Austin, TX

June 2021 – July 2022

Senior Subject Matter Expert

Served as escalation point and systems thinker across moderation operations, bridging frontline reviewers, QA, and leadership on complex policy decisions.

  • Led a team of 7 SMEs supporting complex moderation cases and technical issue resolution
  • Developed knowledge bases and decision frameworks used by 100+ moderators, reducing handle time by 20%
  • Trained teams on gray-area enforcement and risk-based decision-making to reduce escalation dependency
  • Implemented performance recovery frameworks for underperforming agents, improving QA scores and retention
  • Conducted root-cause analysis on trends and abnormalities, identifying automation opportunities and reporting patterns to leadership

Accenture (Meta) – Austin, TX

February 2020 – June 2021

Platform Quality Assurance Specialist

Focused on reviewer accuracy, evaluation consistency, and policy calibration in a fast-moving moderation environment where quality drift could quickly affect enforcement outcomes.

  • Assessed QA error trends with leads to refine evaluation criteria through data-backed calibration
  • Designed manuals, ticketing playbooks, and compliance scoring rubrics to improve review consistency
  • Organized training sessions for 40+ moderators to strengthen structured policy application
  • Supported internal QA audits, onboarding walkthroughs, and biweekly business review documentation
  • Observed that unclear review standards created repeated quality variance across agents and edge-case queues

Accenture (Meta) – Austin, TX

August 2019 – February 2020

Content Moderator Analyst

Worked directly in frontline review queues, where repeated exposure to abuse patterns, reviewer variation, and edge-case content built the foundation for later QA and systems-focused work.

  • Reviewed high-risk content queues while maintaining 100% SLA compliance across enforcement categories
  • Identified emerging abuse patterns and recurring review challenges that informed policy refinement and coaching
  • Tracked reviewer performance data to surface regression risks and coaching needs
  • Shortened ramp-up time for new moderators through training, feedback, and calibration support
  • Built early experience in how policy ambiguity creates inconsistency in real moderation environments

Flagship Case Study

AI Safety Evaluation Under Production Pressure

This case study is based on the same kind of operating pattern I have seen across AI evaluation, moderation QA, and high-risk Trust & Safety work: the system can look healthy in aggregate while the dangerous misses collect in small, ambiguous categories. That is where I spend most of my attention.

Observed Scenario

Prompts and model responses involving minors, sexualized language, grooming indicators, self-harm, or coercion can fail quietly when the review process treats them like ordinary policy calls. The language may be indirect. The risk may sit in context. The correct action may require escalation discipline, not just a label.

Primary risk Under-escalation
Control needed Calibration
Downstream effect Safety exposure

Why It Breaks

Most safety policies are written to be precise, but production cases arrive messy. Reviewers and models can both miss risk when they over-index on explicit terms and underweight context, pattern, age signals, or coercive intent.

Operational Impact

  • High-risk cases can be routed as routine review instead of specialized escalation
  • Evaluator confidence can look high even when the underlying reasoning is thin
  • Legal and regulatory pathways, including NCMEC-related escalation, depend on consistent triage
  • Teams lose time when QA has to reverse decisions late in the process

What This Shows

AI safety evaluation needs category-level controls, not just overall accuracy. For sensitive content, the work is to make the escalation threshold clear enough that careful reviewers and automated systems stop drifting in opposite directions.

Where Risk Hides
Explicit violation
Lower ambiguity
Context-dependent risk
Higher ambiguity
Indirect grooming/coercion signals
Needs escalation logic

Illustrative pattern based on operational review experience, not a published dataset.

Control Design
Weak control Reactive

QA finds the miss after the decision

Strong control Preventive

Guidance flags risk before routing fails

For child safety and self-harm categories, earlier control design matters more than prettier dashboard numbers.

Observed Scenario

Model responses or evaluator decisions can appear policy-compliant at the surface while missing the actual enforcement intent. I saw this most clearly in LLM QA and compliance evaluation work, where a prompt can be answered cleanly but still be unsafe, evasive, or misaligned with the policy objective.

Why It Breaks

The failure is usually not that the policy is absent. The failure is translation. Policy language, prompt design, reviewer rubrics, and model behavior each carry their own assumptions. If nobody reconciles those assumptions, QA becomes late-stage cleanup.

Operational Impact

  • Model-policy alignment appears better than it is because edge cases are averaged away
  • Evaluators apply different reasoning to the same policy standard
  • Prompt engineers receive vague feedback instead of actionable failure patterns
  • Teams spend cycles fixing symptoms instead of the rubric or prompt family causing the drift

What I Would Build

A useful evaluation loop separates the error type before proposing a fix: bad prompt, bad model behavior, unclear policy, weak rubric, evaluator calibration, or escalation threshold. That is how you get from “accuracy is down” to a concrete repair.

Observed Scenario

Fraud, spam/scam, seller abuse, and coordinated manipulation rarely show up as one perfect signal. They show up as weak signals that become meaningful only when clustered: account history, language reuse, timing, seller behavior, payment risk, prior enforcement, and reviewer notes.

Why It Breaks

Queues and dashboards are often optimized for volume. Risk work needs prioritization. When systems treat every signal as equal, the important thing is easy to bury under the merely noisy thing.

Operational Impact

  • Early risk patterns wait too long for confirmation
  • False positives create seller or user friction when signals are not weighted carefully
  • Teams miss the difference between one bad item and a repeatable abuse pattern
  • Policy, QA, and product teams see different fragments of the same problem

What This Shows

Good Trust & Safety operations depend on evidence weighting. The job is not to collect every signal. The job is to decide which signals deserve action, review, escalation, or better instrumentation.

Interactive Tool: Enforcement Tradeoff Simulator

Every safety system has a threshold problem. Push too hard and you create false positives, appeals, reviewer friction, and business impact. Push too softly and harmful content remains active. The work is not pretending the tradeoff is avoidable. The work is knowing which categories cannot tolerate the same threshold.

False Positives
50%

Incorrect enforcement against non-violating content rises as rules get stricter.

False Negatives
50%

Missed harmful content rises as rules get looser.

Reviewer Friction
50%

Friction increases when ambiguous content is pushed through stricter enforcement thresholds.

Platform Exposure
50%

Exposure increases when looser enforcement allows more harmful content to remain active.

How I Work

I tend to look for the control that is missing or too vague. In AI safety and moderation work, that usually means tracing a bad outcome backward until the failure is specific enough to repair.

  • Separate policy failure from evaluator failure, model failure, and workflow failure
  • Use root-cause analysis before adding another training slide or dashboard
  • Build escalation paths that make the safest action the easiest action
  • Give prompt engineers and policy teams evidence they can actually use

That is the perspective I bring to AI evaluation, compliance QA, moderation quality, and Trust & Safety operations.

Additional Case Studies

AI Evaluation Calibration

This case reflects work across LLM QA, red-team prompt review, and vendor quality audits where model output, evaluator reasoning, and policy intent all had to align under production constraints.

Observed Scenario

Evaluators could reach different decisions on the same prompt-response pair because the rubric did not clearly separate harmful content, allowed discussion, evasive model behavior, and escalation-worthy ambiguity.

Primary issue Calibration
Control point Rubric clarity
Failure type Reasoning drift

Why It Breaks

Policies are written to define intent, but evaluation work needs operational examples, boundary cases, and repeatable reasoning. Without that, the same policy becomes several different policies in practice.

QA Correction Timing
Clear rubric examples
Stronger agreement
Gray-area examples
Needs calibration
Operational Effect
Before calibration Variable

Decisions depend on reviewer interpretation

After calibration Stable

Reasoning is easier to audit and coach

Key Takeaway

QA is not just validation. In AI safety work, it is a control system for policy interpretation, evaluator behavior, and model feedback quality.

High-Risk Escalation Design

Based on moderation and compliance work across child safety, sextortion, spam/scam, fraud, and sensitive-content queues where the wrong routing decision can matter more than the initial label.

Observed Scenario

Reviewers often knew a case felt risky but did not always have a clean path for documenting why, escalating it, or distinguishing a policy violation from a legal or safety escalation trigger.

Main risk Routing error
Control needed Decision tree
Failure mode Late escalation

Why It Breaks

Escalation guidance is often written after the policy, when it should be part of the policy design. If reviewers have to invent the path during live work, the system is already taking unnecessary risk.

Escalation Quality
Standard queue label
Not enough
Documented risk rationale
Audit-ready
Impact Comparison
Loose path Slow

Escalation depends on individual judgment

Clear path Fast

Risk is routed with evidence and urgency

Key Takeaway

Strong escalation design is a safety mechanism. It protects users, reviewers, vendors, and the platform by making high-risk decisions consistent enough to defend.

Certifications & Education

Associate of Applied Science (AAS) in Cybersecurity

Austin Community College

Certified ScrumMaster (CSM)

Scrum Alliance

ITIL® 4 Foundation

PeopleCert

Google Data Analytics Professional Certificate

Google / Coursera

AWS Certified Cloud Practitioner

Amazon Web Services

Lean Six Sigma Yellow Belt

CSSC

Lean Six Sigma White Belt

CSSC

Google IT Support Professional Certificate

Google / Coursera

Customer Success Manager (CCSM Level 1)

SuccessCOACHING

Project Management

University of California, Irvine / Coursera

Generative AI for Everyone

DeepLearning.AI

AI Fundamentals

Professional Development

AI Security

Securiti

Selected Highlights

AI Safety Evaluation

Evaluate prompts and model responses for safety, compliance, child-safety risk, self-harm risk, harmful output patterns, and policy alignment across current AI evaluation programs.

High-Risk Operations at Scale

Managed Meta program review teams across high-risk queues, maintained SLA performance, scaled Arabic-language counterterrorism operations, and built decision structures that improved first-pass accuracy.

QA Systems & Vendor Calibration

Built QA loops, evaluator coaching, knowledge bases, dashboards, and vendor review rhythms that turned recurring mistakes into clearer controls and measurable improvement.

Let’s Connect

I’m most interested in AI safety, model evaluation, Trust & Safety, compliance QA, and risk operations roles where policy judgment has to survive real production pressure.

Email Me LinkedIn