Mong'Are N.

Key Skills

Software

CloudFactory

Appen

iMerit

Micro1

Toloka

Internal/Proprietary Tooling

Telus

Mindrift

Top Subject Matter

Satellite imagery analytics and 3D roof model creation

Autonomous robotics training datasets (egocentric vision)

LLM alignment

Top Data Types

Image

Text

Document

Top Task Types

Data Collection

RLHF

Polygon

Fine-tuning

Bounding Box

Segmentation

Cuboid

Polyline

Evaluation/Rating

Freelancer Overview

I am a data analyst and AI training specialist with hands-on experience in data labeling, AI model evaluation, RLHF, DPO, and multimodal AI systems. At CloudFactory, I analyze and interpret satellite imagery to create accurate 3D roof models using spatial data and specialized annotation tools. My work involves detailed geospatial analysis, quality assurance, and precise data structuring to support machine learning workflows. I have also developed strong expertise in edge-case handling, data validation, and maintaining annotation consistency across projects. In my AI training roles at Turing and Micro1, I worked on large-scale AI training and evaluation projects involving RLHF, DPO, reward modeling, multimodal RAG systems, and robotics datasets. I curated and evaluated prompt-response datasets, ranked AI outputs for factuality and coherence, and contributed to benchmark evaluations for vision-language models across multiple domains. I also engineered high-quality egocentric vision datasets for autonomous robotics training, focusing on spatial reasoning, human demonstration data, and complex real-world interactions. My background in engineering, analytical thinking, and technical documentation enables me to deliver high-quality, accurate, and scalable AI training data solutions.

Labeling Experience

Analyst at CloudFactory (April 2024 – Present)

CloudFactory3D SensorPolygon

Analyzed and interpreted satellite imagery to support the creation of accurate 3D roof models of buildings. Modeled roof dimensions, pitch, and edge types based on spatial data to improve training-ready annotations for downstream 3D understanding tasks. Collaborated with the team to streamline the workflow and ensure accurate model rendering quality. • Satellite imagery analysis • 3D roof model attribute modeling (dimensions, pitch, edge types) • Workflow streamlining for efficient rendering • Quality assurance for model accuracy

2024 - Present

AI Robotics Analyst at Micro1 (January 2026 – May 2026)

VideoData Collection

Engineered high-fidelity egocentric (POV) vision datasets to train autonomous household robotics for dexterous tasks. Curated multimodal training assets for multi-stage domestic workflows while adhering to spatial reasoning, lighting, and motion-blur benchmarks for computer vision models. Standardized data acquisition protocols with wide-angle capture systems to optimize signal-to-noise ratio for large world models. • Egocentric POV vision dataset engineering • Multimodal training asset curation with CV benchmarks • Wide-angle capture protocol standardization for SNR • Edge-case documentation for improved robotic perception

2026 - 2026

AI Analyst (AI Engineer) at Turing (October 2024 – November 2025)

TextRLHF

Designed DPO and RLHF pipelines using pairwise preference ranking on chosen/rejected response pairs to reduce alignment training time. Built large-scale prompt-response datasets and trained reward models to predict human preference signals for factuality, coherence, and helpfulness. Fine-tuned multimodal RAG agents and vision-language models and contributed to benchmark evaluation using LLM-as-a-judge scoring. • Pairwise preference ranking (chosen vs rejected) for RLHF/DPO • Prompt-response dataset construction • Reward model training for preference prediction • Multimodal RAG/VLM fine-tuning and evaluation scoring

2024 - 2025

Education

Work History