Baracka K.

Key Skills

Software

Don't disclose

AWS SageMaker

V7 Labs

Top Subject Matter

LLM output evaluation and financial record structuring

LLM workflow parsing

cost extraction

Top Data Types

Text

Document

Image

Top Task Types

Function Calling

Fine-tuning

Classification

Polygon

Cuboid

Segmentation

Freelancer Overview

AI Output Evaluation & Data Structuring — ChamaLedger Project. Core strengths include Don't disclose, Internal, and Proprietary Tooling. Education includes Diploma in Computer Science, Zetech University (2026). AI-training focus includes data types such as Text and labeling workflows including Evaluation, Rating, and Function Calling.

Labeling Experience

AWS AI/ML Scholars Program — Amazon Web Services (2026–Present)

AWS SageMakerTextFine-tuning

Participated in hands-on AWS-sponsored training focused on ML model training, NLP foundations, and responsible AI practices. Completed work involving data preprocessing, model evaluation, and deployment to better understand training data quality from the model’s perspective. Applied learning relevant to creating high-quality datasets for downstream NLP and evaluation tasks. • Trained on NLP-oriented ML concepts and workflows • Practiced data preprocessing and evaluation procedures • Studied responsible AI considerations for model behavior • Gained understanding of what constitutes good training data

2026 - Present

LLM API Response Analysis — SemanticGate Library

TextFunction Calling

Built and deployed parsers for OpenAI and Anthropic API response formats, including token and metadata-aware structured parsing. Analyzed response fields to extract and categorize costs for classification and budget tracking, effectively performing structured annotation of API outputs. Developed admin tools to define and update annotation-style policies such as rate rules and cost thresholds. • Parsed OpenAI/Anthropic response formats with token-count and metadata analysis • Implemented cost extractors for categorization and budget tracking • Created workflow/admin tooling for annotation-style policy updates • Supported robust LLM evaluation pipelines using structured JSON outputs

2024 - 2024

AI Output Evaluation & Data Structuring — ChamaLedger Project

Don't discloseText

Evaluated AI-generated financial record outputs extracted from unstructured WhatsApp conversations and validated them against source data to ensure accuracy. Structured the resulting information into clean records using a consistent schema and maintained an append-only audit ledger for traceability. Focused on detecting hallucinated figures and misclassified entries while ensuring guideline adherence across the dataset. • Cross-referenced GPT-4o extractions against original conversation content • Corrected model errors, hallucinations, and classification mistakes • Organized data into contributions, loan entries, and action items • Ensured full traceability through an append-only ledger

2024 - 2024

Education

Work History