Parikshit Parihar

Key Skills

Software

Roboflow

Label Studio

Scale AI

CVAT

Labelbox

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Audio

Top Task Types

Bounding Box

Segmentation

Data Collection

Point Key Point

Tracking

Emotion Recognition

Text Summarization

Fine Tuning

Evaluation Rating

Prompt Response Writing SFT

Audio Recording

Freelancer Overview

I am an AI and Data Engineer with over 3 years of hands-on experience specializing in data labeling and annotation for computer vision applications, particularly in safety-critical domains like surveillance and PPE detection. My expertise includes curating large-scale video datasets, defining and enforcing annotation guidelines, and building semi-automatic pipelines using tools such as CVAT, Roboflow, and Label Studio to improve efficiency and data quality. I have a strong track record in stress-testing AI models against real-world edge cases—such as shifting lighting, occlusions, and moving shadows—and leveraging performance metrics like accuracy, precision, recall, and confusion matrices to drive significant model improvements. I am skilled at collaborating with cross-functional teams, validating ground truth data, and ensuring robust, reliable training datasets that align with client and product needs.

Labeling Experience

Speech Detection

LabelboxAudioData CollectionPrompt Response Writing SFT

Currently Labelling speech datasets for AI voice systems, covering speech segmentation, pronunciation accuracy, intonation, and naturalness. Worked with diverse accents, speaking styles, and background noise to support training of natural-sounding, human-like AI speech models.

Currently Labelling speech datasets for AI voice systems, covering speech segmentation, pronunciation accuracy, intonation, and naturalness. Worked with diverse accents, speaking styles, and background noise to support training of natural-sounding, human-like AI speech models.

2025

LLM training

LabelboxTextText SummarizationFine Tuning

Currently preparing and curating high-quality text datasets for LLM training and fine-tuning, including data cleaning, instruction formatting, prompt–response alignment, and quality validation. Ensured consistency, reduced bias and noise, and improved model performance through iterative dataset refinement.

Currently preparing and curating high-quality text datasets for LLM training and fine-tuning, including data cleaning, instruction formatting, prompt–response alignment, and quality validation. Ensured consistency, reduced bias and noise, and improved model performance through iterative dataset refinement.

2025

Sentiment Analysis

RoboflowTextEmotion RecognitionText Summarization

Annotated large volumes of Twitter/X text data for sentiment analysis (positive, negative, neutral). Handled slang, sarcasm, emojis, and multilingual content. Ensured labeling consistency through predefined sentiment guidelines and validation checks.

Annotated large volumes of Twitter/X text data for sentiment analysis (positive, negative, neutral). Handled slang, sarcasm, emojis, and multilingual content. Ensured labeling consistency through predefined sentiment guidelines and validation checks.

2025 - 2025

Slip and Fall Detection

CVATImagePoint Key PointTracking

Annotated video frames for slip and fall detection, labeling human postures and fall events. Addressed edge cases such as sitting vs falling, partial occlusions, and varied camera perspectives to ensure accurate event classification.

Annotated video frames for slip and fall detection, labeling human postures and fall events. Addressed edge cases such as sitting vs falling, partial occlusions, and varied camera perspectives to ensure accurate event classification.

2025 - 2025

Fire and Smoke Detection

Scale AIImageBounding BoxSegmentation

Annotated images and video frames for fire and smoke detection in industrial and commercial environments. Handled challenges such as small fire sources, smoke vs fog/dust, low-light conditions, and varied camera angles. Ensured high annotation accuracy through strict guidelines and quality checks.

Annotated images and video frames for fire and smoke detection in industrial and commercial environments. Handled challenges such as small fire sources, smoke vs fog/dust, low-light conditions, and varied camera angles. Ensured high annotation accuracy through strict guidelines and quality checks.

2024 - 2025

Show more

Education

Work History