Laurent Plisson

Key Skills

Software

OneForma

Telus

Other

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Document

Video

Top Task Types

Classification

Data Collection

Diagnosis

Evaluation Rating

Question Answering

Freelancer Overview

I have extensive experience in data labeling and AI training for content moderation, toxicity detection, text classification, and LLM evaluation, with a focus on both English and French. My work includes annotating and curating high-quality datasets for sensitive tasks such as hate speech detection, harassment filtering, and policy violation classification across social platforms and communication tools. I specialize in fine-grained multi-label annotation, context-sensitive classification, and applying nuanced linguistic judgment—crucial for handling ambiguous or borderline cases in both languages. My bilingual fluency ensures consistent and culturally aware labeling, especially important in diverse moderation contexts. In addition to labeling, I’ve contributed to the evaluation and tuning of large language models (LLMs) by conducting human-in-the-loop evaluations, such as ranking outputs, assessing helpfulness, detecting hallucinations, and verifying factual accuracy. This includes preference modeling and instruction-following assessment, where I’ve provided structured, detailed feedback in both English and French. What sets me apart is a combination of linguistic precision, deep familiarity with content policy guidelines, and experience with LLM behaviors, enabling me to produce training data that directly improves safety, reliability, and multilingual performance of AI systems.

Labeling Experience

French/English content moderator

TelusVideoEntity Ner ClassificationClassification

Objective : To detect, flag, or remove violating, harmful, or policy-breaking content in video and audio formats. This involves reviewing and labeling multimedia content based on platform-specific guidelines (e.g., Meta) or legal requirements. Challenges: Multi-modal signals (visual, audio, text/subtitles) Contextual interpretation (intent, sarcasm, background noise)Multilingual and cultural sensitivity High data volume and manual review time Specific Data Labeling Tasks, Scene classification Label types of scenes: violence, adult content, drug use, etc. Temporal labeling (time-stamped annotations) Mark start and end times of violating content within videos. Object & action detection Identify weapons, nudity, blood, fights, or illegal symbols. Contextual flagging, Differentiate between real vs. simulated (e.g., movie scenes vs. actual violence). Text-in-video labeling. Annotate harmful or inappropriate text (e.g., subtitles, signs). Multi-label tagging, A single video might be labeled.

Objective : To detect, flag, or remove violating, harmful, or policy-breaking content in video and audio formats. This involves reviewing and labeling multimedia content based on platform-specific guidelines (e.g., Meta) or legal requirements. Challenges: Multi-modal signals (visual, audio, text/subtitles) Contextual interpretation (intent, sarcasm, background noise)Multilingual and cultural sensitivity High data volume and manual review time Specific Data Labeling Tasks, Scene classification Label types of scenes: violence, adult content, drug use, etc. Temporal labeling (time-stamped annotations) Mark start and end times of violating content within videos. Object & action detection Identify weapons, nudity, blood, fights, or illegal symbols. Contextual flagging, Differentiate between real vs. simulated (e.g., movie scenes vs. actual violence). Text-in-video labeling. Annotate harmful or inappropriate text (e.g., subtitles, signs). Multi-label tagging, A single video might be labeled.

2022 - 2025

Education

Work History