Ervina Laraswati

Evaluating conversational AI outputs for human-like quality and usefulness. Comparing model
responses and providing detailed quality assessments. Audio data annotation and transcription. Testing AI understanding in noisy and real-world speaking environments. Working on multilingual (English & Indonesian) AI Training and localization-related tasks.

Evaluated AI-generated conversations and responses for naturalness, coherence, grammar, accuracy, and human-like quality. Performed multilingual QA, content review, and AI training tasks in Indonesian and English. Handled various AI-related projects, including data annotation and PDF-to-JSON data extraction. Adapted to diverse project requirements while maintaining quality and
consistency.

Developed and refined training data for large language models through content evaluation, response ranking, annotation, and quality review. Assessed AI-generated outputs for accuracy, helpfulness,
naturalness, and instruction-following capabilities. Conducted safety and content evaluations,
including harmful and NSFW content assessment, to support model alignment and responsible AI
behavior.

Worked on quality assurance for AI-generated translations, reviewing and refining English–Indonesian outputs to ensure grammatical accuracy, natural fluency, and cultural relevance. This role involved post-editing machine translations, providing structured feedback, and delivering high-quality results under tight deadlines to support the improvement of AI translation models.

Created conversational scripts for AI chat applications, ensuring responses were natural, contextually relevant, and user-friendly. Worked on refining chatbot interactions to improve engagement and usability.

Transcribed and reviewed audio content with high accuracy using Appen’s tools. Performed speaker
diarization by identifying and tagging multiple speakers in audio files. Annotated additional features such as background noise, pauses, and other relevant audio elements. Ensured compliance with project guidelines and maintained quality standards for transcription and annotation tasks.

Annotated e-commerce datasets through image and text labeling, object classification, and product categorization. Performed image annotation tasks including bounding box creation, object identification, and visual data labeling to support machine learning model training. Reviewed
annotation quality to ensure accuracy, consistency, and compliance with project guidelines.

Evaluated and improved AI-generated content through annotation, response assessment, and quality review. Reviewed outputs for linguistic accuracy, cultural relevance, coherence, and instruction adherence. Performed safety-focused evaluations, including harmful content and NSFW assessments, to support model alignment and content policy compliance.

Translated various academic, legal, and business documents, including thesis abstracts and corporate reports. Provided subtitling services for short films (e.g. Indonesian/Javanese to English or vice versa) for a YouTube release. Performed manual timecoding using Aegisub to ensure accurate subtitle synchronization. Maintained consistency in tone, register, and cultural nuances for target audience.

Designed dynamic dialogues for NPCs in interactive storytelling projects. Focused on character consistency, narrative depth, and engaging player interactions. Developed branching dialogue structures to enhance user experience.

Activities and societies: Volunteering in public activities regarding mental health, Sponsorship Team for English Language Education department events, Committee member in Inclusion Festival, Reporter and Editor at Student Press Agency.

Studied in the English Education Program, engaging in various volunteering activities related to mental health and participating in departmental events. Contributed as a committee member for the Inclusion Festival and worked as a reporter and editor at the Student Press Agency.