Omer Cuvan

Key Skills

Software

Scale AI

Label Studio

Doccano

Top Subject Matter

No subject matter listed

Top Data Types

Text

Top Label Types

Classification

Evaluation Rating

Prompt Response Writing SFT

Computer Programming Coding

Freelancer Overview

I have over 9 years of experience as a software engineer specializing in Python-based systems, AI/ML integrations, and the validation of algorithmic outputs. My work has focused on evaluating and improving AI-generated solutions by validating numerical results with tools like NumPy, SciPy, and Pandas, and providing structured technical feedback to refine model behavior. I have hands-on experience in prompt refinement, code review, and establishing scoring criteria for multi-step problem-solving, ensuring that AI outputs align with first-principles reasoning and accepted standards. My background includes building and validating AI-powered features for virtual assistants, developing backend services and APIs, and supporting benchmarking and evaluation workflows in both technical and collaborative remote environments. I am skilled at ensuring data quality, annotating and validating data flows, and documenting edge cases to support high-quality AI training data.

Labeling Experience

Prompt + Response Authoring for SFT Data (Task & Scheduling)

Label StudioTextPrompt Response Writing SFT

Wrote high-quality prompts and gold-standard responses for multi-step task workflows (scheduling, prioritization, constraints, clarifications). Ensured responses followed first-principles reasoning, used explicit assumptions, and adhered to formatting/structure requirements. Reviewed and corrected model drafts to produce consistent SFT training examples.

Wrote high-quality prompts and gold-standard responses for multi-step task workflows (scheduling, prioritization, constraints, clarifications). Ensured responses followed first-principles reasoning, used explicit assumptions, and adhered to formatting/structure requirements. Reviewed and corrected model drafts to produce consistent SFT training examples.

2023 - 2025

LLM Output Evaluation & Algorithmic Validation for AI Assistant

Scale AITextClassificationEvaluation Rating

Evaluated AI-generated multi-step reasoning outputs for correctness, logical consistency, and adherence to first-principles engineering standards. Performed structured scoring of model responses, validated numerical and algorithmic outputs using Python (NumPy, SciPy, Pandas), and documented edge cases. Designed evaluation rubrics, conducted regression testing, and provided prompt refinement feedback to improve model reliability and reduce hallucinations. Reviewed thousands of model outputs across scheduling, task optimization, and structured problem-solving workflows. Ensured consistency between backend validation logic and model responses under real-world edge cases.

Evaluated AI-generated multi-step reasoning outputs for correctness, logical consistency, and adherence to first-principles engineering standards. Performed structured scoring of model responses, validated numerical and algorithmic outputs using Python (NumPy, SciPy, Pandas), and documented edge cases. Designed evaluation rubrics, conducted regression testing, and provided prompt refinement feedback to improve model reliability and reduce hallucinations. Reviewed thousands of model outputs across scheduling, task optimization, and structured problem-solving workflows. Ensured consistency between backend validation logic and model responses under real-world edge cases.

2023 - 2025

Code & Algorithm Correctness Labeling (Backend + Logic QA)

DoccanoTextEvaluation RatingComputer Programming Coding

Evaluated code and algorithmic solutions for correctness, constraints, and edge-case handling. Labeled outputs as pass/fail with root-cause tags (incorrect assumptions, boundary errors, missing checks, performance issues). Verified results via unit tests and Python-based reproduction for deterministic validation.

Evaluated code and algorithmic solutions for correctness, constraints, and edge-case handling. Labeled outputs as pass/fail with root-cause tags (incorrect assumptions, boundary errors, missing checks, performance issues). Verified results via unit tests and Python-based reproduction for deterministic validation.

2018 - 2025

Support Automation QA Labeling

Label StudioTextClassificationEvaluation Rating

Labeled automation outcomes for customer support workflows (tagging, routing, triggers) and validated rule behavior against expected logic. Classified failures by category (wrong routing, missing conditions, incorrect state transitions) and documented reproducible cases for engineering/QA.

Labeled automation outcomes for customer support workflows (tagging, routing, triggers) and validated rule behavior against expected logic. Classified failures by category (wrong routing, missing conditions, incorrect state transitions) and documented reproducible cases for engineering/QA.

2020 - 2023

Education

Work History