Flashkaysta O.

Key Skills

Software

Other

Top Subject Matter

Neurosurgery / Medical imaging (glioma segmentation)

AI fact-checking / LLM verification

Wearable tech / Emotional and biometric analysis

Top Data Types

Text

Audio

Video

Top Task Types

Segmentation

Question Answering

Emotion Recognition

Function Calling

RLHF

Freelancer Overview

NeuralEdge AI — Glioma Segmentation (project). Core strengths include Other. Education includes Bachelor of Education, Ho Chi Minh City University of Education. AI-training focus includes data types such as Medical, DICOM, and Text and labeling workflows including Segmentation, Question Answering, and Emotion Recognition.

Labeling Experience

RAG-powered AI verification system

OtherTextFunction Calling

Developed an RAG-powered AI verification system that required labeling of retrieval outputs and verification judgments. Annotation focused on mapping claims to relevant sources and labeling which retrieved passages support or refute statements. The resulting labeled data was used to improve retrieval effectiveness and verification correctness. • Labeled claim-to-evidence relevance • Tagged verification outcomes (supported/refuted/uncertain) • Curated retrieval passages for training • Supported evaluation of RAG-based verification responses

2026 - 2026

Real-Time AI Fact-Checking Agent

OtherTextQuestion Answering

Built or prototyped an AI real-time fact-checking system that relies on structured text inputs and generated verifications. The approach required preparing labeled prompt–response pairs and/or verification outcomes to guide model behavior. Annotation efforts focused on correctness, evidence alignment, and classification of verified versus unverified claims. • Labeled claims and corresponding verification results • Curated evidence alignment annotations • Prepared training examples for fact-checking behavior • Supported evaluation of response correctness

2026 - 2026

NeuralEdge AI — Glioma Segmentation (project)

OtherSegmentation

Designed and developed a deep-learning pipeline for glioma segmentation to support precision neurosurgery workflows. The work involved producing segmentation labels aligned with tumor and margin regions for model training and evaluation. Data preparation and annotation consistency checks were used to improve training reliability and inference quality. • Glioma margin segmentation dataset labeling • Label quality control and consistency verification • Training-ready dataset preparation for deep learning • Evaluation support for segmentation model outputs

2026 - 2026

LumiVoice — Wearable tech for real-time biometric and emotional voice analysis

OtherAudioEmotion Recognition

Worked on wearable-based systems intended to infer emotional or biometric states from human signals. The labeling effort likely included assigning emotion/emotional-state labels to synchronized wearable observations for training and evaluation. Annotation supported downstream modeling for real-time analysis and recognition. • Emotion-state labeling for wearable sensor observations • Alignment of labels with time windows • Dataset preparation for emotion recognition models • Supporting model evaluation with annotated outcomes

2024 - 2024

Education