Marshid J.

Key Skills

Software

Other

Remotasks

Top Subject Matter

Fintech credit risk scoring using M-Pesa transaction data

Health triage conversational AI and medical query handling

General AI content evaluation for math/finance/reasoning

Top Data Types

Text

Top Task Types

RLHF

Prompt + Response Writing (SFT)

Entity (NER) Classification

Freelancer Overview

Founder & Lead Product Developer — CreditIQ (AI-powered alternative credit scoring). Brings 3+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Internal, Proprietary Tooling, and Other. Education includes Bachelor of Commerce, KCA University (2028) and CPA Qualification, ICPAK — Institute of Certified Public Accountants of Kenya (2024). AI-training focus includes data types such as Text, Medical, and DICOM and labeling workflows including RLHF, Prompt + Response Writing (SFT), and Evaluation.

Labeling Experience

AI Training Contributor — Outlier AI / Remotasks

RemotasksText

Completed AI training tasks involving rating, ranking, and correcting AI-generated content across math, finance, and general reasoning domains. Applied detailed rubrics to judge factual accuracy, coherence, and instruction-following quality. Maintained high accuracy by following quality-controlled batch instructions and annotation guidelines. • Rated and ranked model responses using rubric-based scoring • Corrected outputs to improve response quality • Evaluated factuality, coherence, and adherence to prompts/instructions • Followed guideline-driven QA processes for consistent dataset labeling

2025 - Present

AI Concept Developer - MedaAI

RemotasksRLHFEntity (NER) Classification

Designed and developed conversational AI flows for a WhatsApp-based health triage product. Created intent-response structures and worked on accuracy and safety evaluation to ensure appropriate outputs for medical scenarios. Established consistent labeling guidelines to support reliable training and improvement of the triage assistant. • Built conversational AI flows for health triage assistant experiences • Evaluated responses for accuracy, safety, and tone using structured rubrics • Developed guidelines for ambiguous health queries to improve consistency • Coordinated iterative improvements to intents and response behaviors

2024 - Present

Founder & Lead Product Developer - CreditIQ

RemotasksTextRLHFEntity (NER) Classification

Founded and leads development of an AI-powered alternative credit scoring platform using M-Pesa transaction data. Designed, evaluated, and iterated on model logic to improve the quality and reliability of credit risk outputs. Built production-ready components for data processing and dashboards, applying fintech and data validation best practices. • Developed and tested an alternative credit scoring model using transaction datasets • Created structured evaluation criteria and prompt-based assessment workflows • Identified bias patterns and edge cases across diverse borrower profiles • Built an M-Pesa statement analyzer and lender dashboard using React and Firebase

2024 - Present

AI Concept Developer — MedaAI (WhatsApp Health Triage Tool)

OtherPrompt + Response Writing (SFT)

Built conversational AI flows for a WhatsApp health triage assistant, requiring intent-response construction and health-query annotation logic. Evaluated AI-generated responses for accuracy, safety, and tone using structured quality rubrics. Developed guidance for ambiguous medical queries to support consistent training outputs. • Authored intent-response pairs and annotated health-related query handling rules • Rated/evaluated response quality against medical rubrics (accuracy, safety, tone) • Produced labeling guidelines for ambiguous triage scenarios • Ensured consistency of annotations to improve model training signal

2024 - Present

Founder & Lead Product Developer — CreditIQ (AI-powered alternative credit scoring)

TextRLHF

Designed and tested an AI credit-scoring workflow that required structured evaluation of AI-generated credit risk outputs. Authored prompt and evaluation criteria to assess outputs in a process similar to RLHF-style annotation. Iterated labeling guidelines by analyzing bias patterns and edge cases across borrower profiles. • Labeled/annotated credit risk evaluation decisions based on model outputs • Wrote structured prompts and evaluation rubrics for consistent scoring • Performed edge-case identification for diverse borrower groups • Validated and improved dataset quality for fintech model training

2024 - Present

Education

Work History