Kendi N.

Key Skills

Software

AWS SageMaker

Anno-Mage

Appen

Argilla

Axiom AI

Clickworker

iMerit

Humanatic

OneForma

Micro1

Mercor

Remotasks

Top Subject Matter

No subject matter listed

Top Data Types

Image

Computer Code Programming

Audio

Top Task Types

Evaluation/Rating

Computer Programming/Coding

Transcription

Freelancer Overview

ANN NKONGE Nairobi, Kenya | [+254110297386] | [[email protected]] | github.com/kendi-dev Professional Profile Telecommunications and Information Engineering graduate with hands-on experience in data systems, backend development, and AI evaluation work. Skilled at working with structured data, classifying conversational outputs, and applying consistent labeling frameworks. Prior AI evaluation roles at Mercor and micro1 have built a strong foundation in intent classification, response quality assessment, and structured JSON annotation directly applicable to financial dataset annotation and function-call tagging. Key Skills Finance & Data Annotation: Intent classification, transaction categorization, merchant labeling, response tone & completeness assessment Function-Call Tagging: Structured JSON records, API-style argument annotation (get_user_summary, get_transactions, run_custom_sql) SQL Literacy: Safe vs. unsafe query identification, parameterized queries, read-only query validation Backend & Data Systems: Laravel, Node.js, PostgreSQL, FastAPI, Docker AI Evaluation: Chatbot response labeling, prompt quality review, structured output validation (Mercor, micro1) Tools: JSON, REST APIs, Git, Grafana/Prometheus, Linux Relevant Experience AI Evaluator / Data Reviewer | Mercor & micro1 | 2024 – Present Evaluated and labeled AI-generated responses for quality, accuracy, tone, and completeness using structured rubrics. Classified intents and flagged incorrect or unsafe

Labeling Experience

AI Chatbot Response Evaluation & Multi-Agent Benchmark Design

TextEvaluation/Rating

Designed and executed structured evaluation tasks for large language model (LLM) outputs as part of Turing's SwarmBench multi-agent benchmark assessment. Work involved rating AI-generated responses across dimensions including accuracy, instruction-following, reasoning quality, and task completion. Produced Harbor-format benchmark submissions defining agent tasks, evaluation criteria, and expected outputs. Additionally completed technical content review tasks assessing code correctness and response quality for Turing's AI training pipeline. Tools used: JSON, structured rubrics, multi-agent task design frameworks.

2026 - 2026

Education