Gordon Omondi Odhiambo

Key Skills

Software

Appen

Clickworker

CloudFactory

Labelbox

Mercor

Mindrift

Remotasks

Toloka

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Geospatial Tiled Imagery

Image

Video

Top Task Types

Data Collection

Evaluation Rating

Land Cover Classification

Mapping

Prompt Response Writing SFT

Freelancer Overview

PhD in Applied Mathematics (UW) and 19 months of specialized training in AI and data annotation in Outlier, Handshake AI, and Remotask. I provide 99.5% of perfect labeling of mathematical expressions, LaTeX, graphs, and video frames with equations or simulations - perfect OCR, math reasoning, and STEM AI models. At Outlier, I was in the top 5 percent of the world by creating high-quality datasets of differential equations, optimization, and stochastic processes. I designed prompts, tested the results of the LLM on mathematical correctness, and created Python validation programs. I obtained 98% inter-rater reliability when rating proofs and dynamical systems answers at Handshake AI. On scale AI thousands of math diagrams have bounding boxes and semantic tags annotated, and I am a teacher (300+ students, 92% pass rate in Calculus), and I make sure the annotations are pedagogically correct, that the AI must learn which mistakes students make. I combine the knowledge of the fields (nonlinear PDEs, epidemiology modeling) and technical excellence (SymPy, Jupyter, LaTeX) and produce training data of publishable quality....

Labeling Experience

LLM Training Data for Advanced Mathematical Reasoning

Scale AIComputer Code ProgrammingQuestion AnsweringText Generation

Produced superior-quality prompt-response pairs of the differential equations, optimization and stochastic processes. Assessed alphabetical output of LLM as accurate and readable. Ranked in top 5% globally. West Wrote Python programs to autovalue symbolic solutions. Chaos theory and epidemiology model generated edge cases.

Produced superior-quality prompt-response pairs of the differential equations, optimization and stochastic processes. Assessed alphabetical output of LLM as accurate and readable. Ranked in top 5% globally. West Wrote Python programs to autovalue symbolic solutions. Chaos theory and epidemiology model generated edge cases.

2024 - 2024

Mathematical Equation & Diagram Annotation for OCR

Scale AIImageBounding BoxEntity Ner Classification

Marked 10,000+ mathematical formulae, graph, and diagrams in scanned textbooks and research papers. Draws variable bounding boxes on equations, variables, and on plots. Training ground truth data on OCR. The highest accuracy achieved was 99.5% in 8 months. Added to sets that were used in actual equation recognition systems.

Marked 10,000+ mathematical formulae, graph, and diagrams in scanned textbooks and research papers. Draws variable bounding boxes on equations, variables, and on plots. Training ground truth data on OCR. The highest accuracy achieved was 99.5% in 8 months. Added to sets that were used in actual equation recognition systems.

2023 - 2023

Education

Work History