Ayush Patel
About
Detail
AI/ML Engineer at Kinaxis
Gujarat, India
Résumé
Jobs
verified_user
0% verified
AI/ML EngineerKinaxisDec 2024 - Current (1 year 9 months)• Improved forecasting accuracy by 28% using TensorFlow and PyTorch, reducing inventory costs and enhancing global supply chain efficiency. • Reduced data latency by 35% automating preprocessing with Python, SQL, and Pandas on AWS, enabling faster retraining cycles. • Deployed Flask and FastAPI APIs on GCP, cutting inference time from hours to seconds for predictive analytics. • Implemented CI/CD pipelines with GitHub Actions, ensuring 100% automated testing and zero-downtime model deployments. • Delivered explainable AI solutions with SHAP and bias detection, ensuring compliance and building stakeholder trust. • Built visual dashboards using Matplotlib and Seaborn, translating ML predictions into actionable insights for executives.- AI EngineerApplyboardMar 2024 - Nov 2024 (9 months)• Increased student-university match accuracy by 25% with recommendation and classification models built in Scikit-learn and PyTorch. • Improved sentiment classification precision by 18% through NLP pipelines developed in TensorFlow and PyTorch. • Maintained 99.9% uptime for 20,000+ daily users by deploying ML services with FastAPI, Docker, and AWS. • Reduced ETL runtime by 30% optimizing data workflows using SQL and Pandas. • Boosted enrollment outcomes by aligning ML insights with admissions strategies across teams. • Developed interactive dashboards to present predictions, accelerating advisor decision-making.
- Machine Learning EngineerKorcomptenz Inc Total Technology TransformationFeb 2022 - Aug 2023 (1 year 7 months)• Enhanced predictive accuracy by 20% building regression, classification, and clustering models in Scikit-learn. • Reduced data preparation time by 50% streamlining preprocessing with SQL and Pandas pipelines. • Deployed scalable ML solutions with Flask APIs on AWS and Azure. • Accelerated client decision-making by 40% with visual insights using Matplotlib and Seaborn. • Reduced defect rate by 30% by applying Git, modular design, and automated testing. • Validated model reliability through statistical analysis, probability testing, and linear algebra methods.
Education
verified_user
0% verified
- Master's degree, Data AnalyticsNortheastern UniversityJan 2023 - Dec 2025 (3 years)
- Bachelor's degree, Computer ScienceIndus UniversityJan 2019 - Dec 2023 (5 years)
Projects (professional or personal)
verified_user
0% verified
- FFake News Detection• Built a machine learning model using Scikit-learn and NLTK to classify 20,000+ news articles, achieving 92% accuracy in distinguishing fake versus authentic news. • Engineered NLP pipelines for tokenization, stop-word removal, and TF-IDF vectorization, improving classification precision and recall while enabling scalable text data preprocessing in Python. • Integrated SQL workflows and visualized patterns with Wordcloud, uncovering misleading vocabulary trends and enhancing model explainability for stakeholders.