Karthik A.

Key Skills

Software

Don't disclose

Top Subject Matter

NLP dataset curation and RLHF evaluation

Computer vision (plant disease detection)

Sentiment analysis and conversational routing dataset preparation

Top Data Types

Text

Image

Video

Top Task Types

Bounding Box

Classification

Prompt + Response Writing (SFT)

Freelancer Overview

Data Labeling & Annotation Specialist — Viswam AI (Applied AI Team) (May 2025 – Present). Brings 2+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Weights & Biases, Don't disclose, and Hugging Face. Education includes Bachelor of Technology, Jawaharlal Nehru Technological University (JNTUH) (2025). AI-training focus includes data types such as Text, Image, and Video and labeling workflows including Evaluation, Rating, and Bounding Box.

Labeling Experience

Data Labeling & Annotation Specialist — Viswam AI (Applied AI Team) (May 2025 – Present)

Text

Curated, manually annotated, and preprocessed 100K+ Indic-language text samples, applying deduplication and quality control to improve downstream model performance. Developed standardized labeling guidelines to ensure consistent annotations across a multi-annotator team. Conducted RLHF evaluation by assessing GPT-style decoder model outputs and providing detailed feedback aligned to human preferences. • Curated 100K+ Indic text samples with quality assurance and deduplication. • Built and rolled out standardized annotation guidelines for team consistency. • Performed RLHF evaluation for GPT-style decoder responses using human preference alignment. • Maintained dataset versioning and logged data quality metrics for traceability using experiment tracking.

2025 - Present

Generative AI Video Annotation & Prompt Curation (Project)

VideoPrompt + Response Writing (SFT)

Maintained a curated collection of scripts and text prompts designed to generate high-quality anime content using AI video tools. Evaluated and categorized outputs from multiple Hugging Face models and online platforms to construct a prompt-to-video evaluation dataset. Structured prompt engineering outputs to support repeatable generation quality assessment. • Maintained prompt/script library for anime content generation. • Evaluated and categorized outputs across multiple models/platforms. • Built a prompt-to-video evaluation dataset. • Supported quality-focused prompt iteration using scripting.

2024 - 2024

Multimodal Sentiment Chatbot — Dataset Preparation (Project)

Don't discloseTextClassification

Labeled and validated a corpus of 160K tweets for sequence classification to build ground-truth data for a BERT model. Annotated conversational data to support routing decisions for dynamic query handling between cloud and local backends. Focused on producing reliable labeled datasets to achieve strong model performance. • Labeled and validated 160K tweets for sequence classification. • Built ground-truth data enabling high BERT F1-score. • Annotated conversational examples for dynamic routing. • Supported dataset preparation and validation for downstream training.

2024 - 2024

Crop Disease Detection — Image Annotation (Project)

Don't discloseImageBounding Box

Annotated and curated a balanced 4-class plant pathology image dataset to produce accurate ground-truth labels for model training and evaluation. Performed quality assurance on augmented images to verify that transformations preserved diagnostic features. Ensured dataset balance and label integrity to support strong downstream classification performance. • Created bounding box/class labels for plant pathology categories. • Curated a balanced 4-class dataset for ResNet training. • Conducted QA checks on augmented datasets to preserve diagnostic features. • Verified ground-truth quality to support high test accuracy.

2024 - 2024

Education

Work History