Maurice Gistach

Key Skills

Software

Appen

Clickworker

CloudFactory

CrowdSource

CVAT

Figure Eight

HiveMind

Humanatic

Lionbridge

Mindrift

OneForma

Remotasks

Scale AI

Toloka

Telus

Other

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Image

Text

Top Task Types

Classification

Computer Programming/Coding

Evaluation/Rating

Prompt + Response Writing (SFT)

RLHF

Freelancer Overview

I have a strong background in AI data annotation, evaluation, and training, gained through working with global platforms such as UHRS, Toloka, Remotasks, Clickworker, CloudFactory, Rev, and DataEQ Crowd among others. My projects have ranged from rating and rewriting GenAI responses to improve accuracy and helpfulness, to sentiment annotation on social media posts, transcription and editing of audio files, search engine and ad relevance evaluation, and even classifying recorded phone calls for motor vehicle companies. This variety gave me the chance to work with text, audio, and digital content, sharpening my ability to adapt quickly and pay close attention to detail." "In every project, I focused on delivering consistent, high-quality results by following clear guidelines, checking my work carefully, and collaborating with others to stay aligned on standards. Whether through calibration sessions, feedback loops, or accuracy reviews, I made sure my contributions improved the reliability of the data and, ultimately, the performance of the AI systems it powered. What sets me apart is not just the range of tasks I’ve handled, but also the value I bring in helping companies build AI assistants, search engines, and customer-facing systems that are more accurate, useful, and trustworthy.

Labeling Experience

Bee Pollination Rating

Scale AITextEvaluation Rating

Dealt with evaluating two responses that were provided as outputs generated by an AI chatbot for a given prompt from a user. Work involved giving feedback on which response was better and why and this was carefully analyzed using a dimensionalized quality metric score that had five factors. These were a.) language mechanics; spelling, grammar etc. b.) Structure and composition; conciseness, formatting and coherence, and tone appropriateness c.) Relevance and completeness d.) Factuality and accuracy e.) Trust and safety

Dealt with evaluating two responses that were provided as outputs generated by an AI chatbot for a given prompt from a user. Work involved giving feedback on which response was better and why and this was carefully analyzed using a dimensionalized quality metric score that had five factors. These were a.) language mechanics; spelling, grammar etc. b.) Structure and composition; conciseness, formatting and coherence, and tone appropriateness c.) Relevance and completeness d.) Factuality and accuracy e.) Trust and safety

2023 - 2024

Bee LLM

Scale AITextPrompt Response Writing SFT

Project dealt with rewriting and rating AI-generated responses to prompts in order to improve the factuality and helpfulness among other qualities of an AI Assistant. The tasks involved reviewing chatbot outputs, identifying gaps or errors, and rewriting them to provide clearer, more accurate, and more complete answers. The project covered a wide range of prompts, from simple everyday queries to complex factual questions, making consistency especially important. To maintain quality, I followed detailed guidelines that cautioned us in order to come up with useful responses, prompts need to be categorized using the following criteria; if a prompt is inapropriate, in foreign language, difficult to follow, identity dependent etc. Once the prompts are categorized, we evaluated the corresponding response and how it addressed the prompt using the following broad criteria; language mechanics, structure and composition, relevance and completeness, factuality and accuracy and trust and safety.

Project dealt with rewriting and rating AI-generated responses to prompts in order to improve the factuality and helpfulness among other qualities of an AI Assistant. The tasks involved reviewing chatbot outputs, identifying gaps or errors, and rewriting them to provide clearer, more accurate, and more complete answers. The project covered a wide range of prompts, from simple everyday queries to complex factual questions, making consistency especially important. To maintain quality, I followed detailed guidelines that cautioned us in order to come up with useful responses, prompts need to be categorized using the following criteria; if a prompt is inapropriate, in foreign language, difficult to follow, identity dependent etc. Once the prompts are categorized, we evaluated the corresponding response and how it addressed the prompt using the following broad criteria; language mechanics, structure and composition, relevance and completeness, factuality and accuracy and trust and safety.

2023 - 2024

Audio filtering

HumanaticAudioClassificationAction Recognition

IProject involved classifying recorded phone call conversations from motor vehicle companies. Each call had to be categorized based on set criteria, such as sales, payments, service, or spare parts, with the goal of ensuring customers were directed to the most suitable outcome. The categories were in ascending order, from basic classfication of inbound or outbound call, to reason for calling al the way to the booking of appointments. Because the recordings came from a variety of clients and scenarios, the task required careful listening and accurate judgment. To maintain quality, I relied on clear classification guidelines and all the calls passed through quality checks to ensure their classification was accurate.

IProject involved classifying recorded phone call conversations from motor vehicle companies. Each call had to be categorized based on set criteria, such as sales, payments, service, or spare parts, with the goal of ensuring customers were directed to the most suitable outcome. The categories were in ascending order, from basic classfication of inbound or outbound call, to reason for calling al the way to the booking of appointments. Because the recordings came from a variety of clients and scenarios, the task required careful listening and accurate judgment. To maintain quality, I relied on clear classification guidelines and all the calls passed through quality checks to ensure their classification was accurate.

2022 - 2024

Sentiment Analysis

OtherTextClassification

Worked on analyzing social media posts to determine the overall sentiment they expressed. Main task was to label each post by choosing the sentiment category that best fit, such as positive, negative, neutral, or mixed. The dataset was fairly large and diverse, which made it important to stay consistent. To ensure quality, I followed clear labelling guidelines, referred to sample examples, and took part in agreement checks with other annotators in the forums. We also had regular feedback sessions and spot checks to make sure the labelling was accurate, consistent, and reliable.

Worked on analyzing social media posts to determine the overall sentiment they expressed. Main task was to label each post by choosing the sentiment category that best fit, such as positive, negative, neutral, or mixed. The dataset was fairly large and diverse, which made it important to stay consistent. To ensure quality, I followed clear labelling guidelines, referred to sample examples, and took part in agreement checks with other annotators in the forums. We also had regular feedback sessions and spot checks to make sure the labelling was accurate, consistent, and reliable.

2021 - 2024

Search engine results/keywords classification

TolokaTextClassification

In this project, I worked on evaluating search engine results to check how well they matched specific keywords. The main task involved reviewing the results returned for given queries and identifying whether they were relevant, accurate, and useful. The project covered a wide range of keywords across different topics, requiring careful judgment to ensure high-quality matches. To maintain consistency and accuracy, I followed clear evaluation guidelines, used rating scales to assess relevance, and answers passed through quality review and grading steps to align with project standards.

In this project, I worked on evaluating search engine results to check how well they matched specific keywords. The main task involved reviewing the results returned for given queries and identifying whether they were relevant, accurate, and useful. The project covered a wide range of keywords across different topics, requiring careful judgment to ensure high-quality matches. To maintain consistency and accuracy, I followed clear evaluation guidelines, used rating scales to assess relevance, and answers passed through quality review and grading steps to align with project standards.

2021 - 2024

Show more

Education

Work History