Hard Parikh

Currently pursuing a Master's in Computer Science from the University of California, Riverside, I am a highly motivated and skilled Computer Science graduate. Proficient in various programming languages and I have hands-on experience with the latest technology. Additionally, I possess expertise in web development and have a strong foundation in Machine Learning frameworks. During my tenure as a Software Developer Engineer at Amadeus Labs, I demonstrated effective teamwork while developing, maintaining, and testing software components.

• Car Type and License Plate Detection (CTLPD):- Real-time image processing to predict not only a car in the given image or video but also the category (Sedan, Hatchback, SUV, Convertible) of the car using OpenCV, Darknet Framework and YOLOv4. Achieved 96.27% confidence for car type and high accuracy for license plate detection. Selected for publication in International Conference on Information and Communication Technology for Intelligent Systems(ICTIS) by Springer.
• People Counting in Dense Crowd:- Researched crowd counting solutions with OpenCV and MobilenetSSD. Collected and preprocessed large-scale datasets from open-source platforms like COCO, PASCAL VOC, and ShanghaiTech. Successfully achieved 92.60% accuracy in crowd counting using YOLOv4, outperforming other competitive models. Accepted for publication in Solid State Phenomena, indexed under Scopus (Q3).
• Neural Machine Translator:- Developed a Machine Translation project focusing on converting human-readable dates, such as '8th of April, 2000,' into machine-readable format '08-04-2000.' Implemented Neural Machine Translation (NMT) approach using a bidirectional recurrent neural network (encoder) to process source sentences into vectors for the decoder to predict words in the target language. Utilized vector representations for words and internal state, enhancing the performance and accuracy of the NMT system compared to traditional methods.
• Song Recommendation System:- Developed a Song Recommendation System with Spotify using content-based filtering approach to predict the next songs of a playlist based on features such as loudness, tempo, energy, and more. Conducted extensive feature engineering, including one-hot encoding and TF-IDF measures, to process metadata and genres for accurate recommendations. Implemented sentiment analysis on track names to extract additional features, enabling the system to recommend songs based on textual information. Utilized cosine similarity measure to find similarities between playlist summarized vectors and non-playlist songs, enabling effective song recommendations for users.

• Machine Learning Specialization by Andrew NG
• Deep Learning Specialization by Andrew NG
• Data Structures and Algorithms Specialization by Michael Levin