Harmish Patel Email & Phone Number

Vaughan, On, Ca

Vaughan, Ontario, Canada

• Led seamless migration from heuristic to automated anomaly detection, utilizing PyTorch Autoencoders and PySpark for image processing, resulting in a 43% increase in workflow productivity.• Implemented dimension and depth models to quantify anomaly size and identify deep features employing a multi-layer neural network and gradient descent with an 80% detection rate.• Built a KPI dashboard to monitor changes and track efficiency and accuracy metrics.• Collaborated with cross-functional teams to integrate detection results into comprehensive pipeline inspection reports.

Windsor, Ontario, Canada

Ontario, Canada

- Helped small and medium entrepreneur through FutureProof program funded by Canadian government.- Guided 6 clients with email marketing, SEO for their online presence, establishing enterprise website for their local business on Shopify and Square & market their products with Google AdSense and Facebook Ads.- Collaborated with other team members for market research, competion research and automate the process using Python script & reduced the number of man hours.

Windsor, Ontario, Canada

● Designed and implemented an end-to-end deep learning architecture-based system for converting handwritten notes images to text using OCR technology, achieving an accuracy rate of ~85%.● Utilized PyTorch pre-trained NLP transformer models to generate abstractive summaries of topics, achieving a ROGUE score of ~55%.

Bengaluru, Karnataka, India

• Deployed a scalable data pipeline on AWS utilizing S3, EMR, and DynamoDB to efficiently extract and load data into the graph database (Neo4j) using batch processing for generating a comprehensive knowledge graph for academic concepts.• Processed a large-scale textbook dataset of 1 TB, transforming it into a model-based format and extracting key entities such as headings, figure names, TF, TF-IDF, and DF using the Pandas and nltk libraries.• Optimized the processing time for mapping the Wikipedia dumps with the extracted entities by 23 hours using a PySpark data pipeline, resulted in a 20% acceleration in data processing speed and a 34% increase in efficiency.• Assigned weights and filtered relevant keywords from the academic corpus, successfully extracting 20,000 potential academic concepts from each chapter.

India

Technical stack: Python, Flask, ReactJS, Time-series analysis, Machine learning- Conducted data scrapping using BeautifulSoup library in Python; fed data to Long Short-Term Memory Machine learning algorithm for time-series prediction for the company’s forex platform; saving time utilized in manual data entry of forex pair prices.

Master'S Degree, Applied Computing

Specialization in Artificial Intelligence

Bachelor Of Technology (B.Tech.), Computer Science

Specialization in Big Data and Analytics