Abdulahi Bello

Data Annotator and AI Trainer with 5+ years of experience delivering high-quality multimodal training data across text, image, video, audio, and 3D for leading AI platforms and Meta projects. Skilled in image, video, and 3D annotation, captioning and timestamping, and AI-training work such as RLHF preference ranking and prompt-and-response evaluation. Also an experienced Machine Learning Developer, building and deploying models in Python, Flask, and scikit-learn.

Native-level English writer who pairs exceptional attention to detail with strong technical literacy and a consistent record of meeting strict quality benchmarks at scale.

Machine Learning Developer at xAI (2025-06 – 2026-02)

• Designed, trained, and evaluated machine learning models in Python (scikit-learn, pandas, NumPy) across 6+ projects spanning classification, regression, and clustering, improving baseline model accuracy by up to 18%.
• Built and deployed 5+ model-serving REST APIs and lightweight web applications with Flask, exposing trained models through clean, documented endpoints.
• Engineered features and ran data cleaning, preprocessing, and hyperparameter tuning, cutting data-preparation time by roughly 30% through reusable pipelines.
• Curated, labeled, and validated 15,000+ training and evaluation records, running benchmarking loops to measure model performance against defined success metrics.
• Documented architecture, experiments, and results to ensure reproducibility and smooth handover across the team.

Data Annotator & AI Trainer at Micro1 (2024-04 – 2025-05)

• Annotated 1,500+ videos and 12,000+ individual frames, labeling objects, actions, and events along the timeline to build training data for computer-vision and video-understanding models.
• Created precise captions and transcriptions for 200+ hours of footage, aligned to timestamps within ±0.2 seconds for subtitle generation and video-language model training.
• Segmented long-form video into labeled clips and annotated start/end times for key moments and scene changes, sustaining a throughput of ~40 clips per day.
• Applied detailed labeling guidelines consistently across 10+ projects, flagged ambiguous and edge-case footage, and proposed rule clarifications that improved dataset quality.
• Conducted quality-assurance reviews of peer annotations, maintaining a 98% acceptance rate and high inter-annotator agreement.

Data Annotator & AI Trainer at Hugotech — Meta Projects (2023-02 – 2024-03)

• Delivered 20,000+ image annotations across 8+ Meta projects — bounding boxes, polygons, semantic segmentation, and keypoint/landmark labeling — at a sustained 350+ images per day.
• Annotated 2,500+ short-form videos (Reels), labeling actions, objects, and content attributes frame by frame to support content-understanding and recommendation systems.
• Produced 3,000+ 3D annotations for Meta using SRT, labeling spatial data to train depth, tracking, and scene-understanding models.
• Maintained 98% labeling accuracy against strict Meta data standards across high-volume annotation batches.
• Performed quality control on completed batches, auditing for labeling errors and keeping the rejection rate below 2%.

Data Annotator & AI Trainer at Mindrift (Toloka AI) (2021-01 – 2023-01)

• Completed 5,000+ multimodal annotation tasks across 15+ projects (text, image, audio, and video) to generate high-quality training data for generative-AI and large language model (LLM) projects.
• Evaluated and ranked 2,000+ prompt-and-response pairs (RLHF-style), rating model outputs for helpfulness, accuracy, and safety.
• Labeled and categorized diverse data types — transcribing audio, tagging images, and classifying text — at 60+ tasks per day under detailed, task-specific rubrics.
• Maintained 97%+ accuracy and reviewed other contributors' work, consistently rated among the platform's top-performing annotators.

B.Eng. in Civil Engineering – University of Ibadan, Nigeria (2021)