Swetha Rayala (she/her/hers)

Education

Phone: +1-919-***-**** Email: ad1ezh@r.postjobfree.com

Linkedin: https://www.linkedin.com/in/swetha-rayala Duke University (Aug'21 - May'23)

MS in Electrical & Computer Engineering (Machine Learning & Data Analytics) Indian Institute of Technology (BHU), Varanasi (July’14-May’19)

Integrated Dual Degree (B.Tech & M.Tech) in Mathematics & Computing Work Experience

Software Engineer - ServiceNow (July'19 - Aug '21) Auto-Routing & Resolution time estimation of Incident Tickets(Classification & Regression)

• Designed and implemented end-to-end training pipelines for various machine learning models, including Logistic Regression, Linear Regression, XGBoost, LIGHTBM, and Feedforward Neural Networks.

• Successfully reduced incident resolution time by 30%, leading to significant cost savings and improved operational efficiency.

• Facilitated more effective resource allocation and ensured compliance with SLAs. Intelligent problem Precedence Identification (Similarity)

• Supported incident resolution by implementing advanced techniques like GLOVE, GUSE, Word2Vec, and TF-IDF to identify similar incidents and adapt to new ones through automated model retraining.

• Customized workflows to handle both Long-Text and Short-Text data effectively. Major Incident Identification, Anomaly Detection and Application Fingerprinting (Clustering)

• Developed training pipelines for clustering algorithms, including K-means, DBScan, and HDBScan.

• Seamlessly integrated with the ServiceNow platform to identify sensitive tickets and enhance IT operations visibility. Python-Based Machine Learning Infrastructure

• Successfully migrated existing Java-based ML solutions from WEKA and Deeplearning4j to Python, leveraging popular libraries like Scikit-learn, PyTorch, and TensorFlow.

• Utilized ONNX for scalable and environment-agnostic model inference. Multilingual Data Processing

• Implemented data processing support for multiple languages using Apache Lucene, including cleaning, stemming, lemmatization, and stopword removal, enhancing machine learning capabilities across languages. Enhanced Scheduling Application for Routing Trainings

• Improved a JavaScript-based scheduling application to optimize the routing of training requests to different regions based on dynamic parameters such as model type, data center location, capacity, and request priority.

• Achieved real-time optimization, resulting in enhanced infrastructure utilization. Streamlined Automated Testing

• Played a pivotal role in the automation of testing processes, contributing to robust unit and integration test environments for both Java and Python ML pipelines.

Development Intern, Root EnRoute (June'22- Aug'22)

Developed robust and fault-tolerant systems for monitoring and rewarding referrers based on user orders.

Successfully implemented a referral program and user wallet, optimizing platform functionality.

Contributed to developing an effective anomaly detection system to improve data accuracy.

Actively participated in the collection and streamlined preparation of data from diverse sources, including internal databases, analytical tools, and CRM systems, to streamline marketing efforts.

Automated essential standard reports for various business areas, enhancing operational efficiency. Technical Skills

Languages: Python, Java, C, C++, SQL

Tools & Technologies: Git, AWS, Terraform, PyTorch, Scikit-learn, JUnit, PyTest, HTML, CSS, Spring, ServiceNow Projects

Improvised Regularization of CNN's Python, PyTorch, Numpy (Nov'22 - Dec'22)

Enhanced convolutional neural networks with advanced regularization techniques: cutout, mixup, and self-supervised rotation prediction.

Conducted experiments on CIFAR10, using these regularization methods as alternatives to standard cross-entropy loss.

Improved ResNet 20 model accuracy by 5% to 8% across diverse test datasets. Sentiment Analysis using generative probabilistic modeling Python, PyTorch, Sklearn, Pandas. (Nov'22 - Dec'22)

Processed and trained the IMDB dataset using Naive Bayes' classifier and LSTM (discriminative Neural Network).

Developed a generative probabilistic model based on the probabilities derived from the Naive Bayes' classifier and evaluated its performance alongside LSTM models on both IMDB data and data generated using the probability model.

Contact this candidate