Harsh S.

Harsh S.

BTech CSE’26 || Former Intern DLJ DRDO || Member of Mensa || GGSIPU || GeeksforGeeks(ADGIPS Chapter) || @Zuno Community Captain

Delhi, India
6roles
13skills
5education
2credentials

About

Proactive and analytical individual with strong logical reasoning and problem-solving skills, eager to learn, collaborate, and contribute to innovative projects.

Experience

Technical Team Member

GeeksforGeeks ADGIPS Chapter

2 yrs 8 mos · Delhi

Organizing Committee Member

Mensa India

4 yrs 5 mos · India

Member

Mensa International

4 yrs 5 mos

Summer Intern

Defence Research and Development Organisation (DRDO)

2 mos

Campus Ambassador

Zuno by foundit

1 yr

Research Intern

Defence Laboratory, Jodhpur (DRDO)

06/2025 – 07/2025 · Jodhpur, India

Built an image analysis program to automatically extract and interpret dominant visual features from varied image inputs; implemented analytical modules to study and visualize patterns in extracted data, enhancing the interpretability of visual content.

Education

Guru Gobind Singh Indraprastha University (GGSIPU), Delhi

Bachelor of Technology - BTech, Computer Science

Oct 2022 - Aug 2026

Indira Ideal Senior Secondary School

Education

Apr 2009 - Mar 2021

Guru Gobind Singh Indraprastha University

B.Tech, Computer Science Engineering

2026

Member of GeeksforGeeks (ADGIPS Chapter); 8.3 CGPA

Indira Ideal Senior Secondary School

12th, PCM

2021

Scored 91.2%

Indira Ideal Senior Secondary School

10th

2019

Scored 85.2%

Skills

Python (Programming Language)Problem SolvingCritical ThinkingPythonC/C++JavaMySQLGoogle Cloud PlatformLooker StudioNeo4jMicrosoft OfficePower BIGitHub

Projects

Deepfake Image Detection

Designed and implemented a custom CNN using Python, TensorFlow, OpenCV, Scikit-learn, and MySQL to classify synthetic facial images. Applied preprocessing and augmentation to improve generalization and reduce overfitting. Achieved 94% training accuracy, 91% validation accuracy, and 91.85% test accuracy, with 0.98 precision for synthetic images and 0.99 recall for genuine faces.

Movie Recommendation System with Sentiment Analysis

Developed a hybrid recommendation engine using content-based filtering and Multinomial Naive Bayes with Python, Flask, Scikit-learn, and the TMDB API. Built a pipeline combining the TMDB 5000 dataset with scraped Wikipedia records, achieving validated sentiment accuracy of 98.77%.