EKTA GUPTA

Full Stack Developer Intern

Amdocs Development Centre Ltd.•  March 2024 - May 2024•  Gurugram, Haryana

Engineered multiple interactive GUI and developed full-stack features for a Generative AI InterviewBot and ChatBot using HTML, CSS, Javascript for frontend and technologies Flask, AJAX and RESTful APIs for Backend which reduces UI load time by 35%. Employed Spacy, OpenAI, NLTK libraries for entity parsing, and applied Prompt Engineering, LangChain, and LLMs for contextual resume analysis and customized question generation, enhancing resume processing accuracy by 50% and generated question relevance by 45%. Utilized Couchbase Server for real-time data handling and JSON database, ensuring scalability and reliability which reduces the manual and administrative workload by 30%.

Summer Intern

Centre of Excellence–AI•  May 2023 - July 2023•  Remote, Delhi

Worked on project ARTGAN which a Generative Adversarial network that can convert real images into artistic canvas and paintings using Deep neural Networks, PyTorch, Keras and Tensorflow for model training, optimisation and fine-tuning of neural networks. The performance metrics of GANs architecture showing Generator model Loss of 0.0655 and Discriminator Loss of 4.3852, indicating high-quality image generation and difficulty for the discriminator to distinguish between real and generated images. Use of NVIDIA GPUs cuda device and T4 GPU runtime which accelerated large-scale computation speed and reduces training time by 75%.

Summer Intern

Centre of Excellence–AI•  May 2022 - July 2022•  Remote, Delhi

Developed a project Human Emotion Detection which can detect Emotions of human face from real-time Video using VS code, haarcascade-frontalface XML file and libraries such as numpy, openCV, tensorflow and keras. Used Python library ImageDataGenerator for image dataset used to reduce overfitting and built CNN on pre-trained model 'MobileNet' which reduced the no. of parameters and computation complexity. It has achieved 83% accuracy and was able to identify most of the emotion of images. Implemented this model in real-time video analysis which was able to classify 4 emotions out of 7 accurately.