Shreyasi Pal

Minneapolis, MN
Phone: 612-666-3218
Email: pal.shreyasi001@gmail.com
Personal Website | GitHub | LinkedIn

Introduction

Machine Learning Engineer with 4 years of production experience, specializing in building high-impact products using ML/AI in collaborative environments.

Skills

Programming Languages: Python, C/C++, R, SQL
Machine Learning Tools: PyTorch, Keras, TensorFlow, OpenCV, SpaCy, Pandas, OpenAI Gym
Cloud Platforms: AWS Lambda, Sagemaker, CDK, Azure App Services

Experience

Delta Air Lines - Enterprise AI Team

AI Engineer
Minneapolis, MN | May 2022 - Present

• Knowledge Management System:
  • GenAI-powered search engine for internal documents used by reservation and customer service agents.
  • Developed Semantic Similarity Query Retriever, Hybrid Search Document Retriever, Retrieval Augmented Generation system with on-the-fly RAG triad filtering and human-in-the-loop feedback system.
  • Used asynchronous implementation, threading, function and LLM caching, and resource provisioning to deliver low latency requirements.
• Net Promoter Score Prediction:
  • Long-term time series prediction of NPS and other satisfaction scores, identifying causal drivers for multiple markets.
  • Developed a transformed-based multivariate time series prediction model.
  • Created causal analysis models for discovering relationships between key variables.
• Automated International Travel Document Verification:
  • Verified travel documents using textual and image features, processing 12.5k documents per country monthly in production.
  • Developed a logo detection and verification algorithm using object detection and matching, text extraction using AWS Textract, and QR code verification.
  • Tools Used: Python, PyTorch, OpenCV, Pandas, Asyncio, Multiprocessing, LangChain, LlamaIndex, AWS Lambda, Sagemaker, CDK.
  • Cloud Tools: AWS Lambda, Sagemaker, CDK.

Pactera Edge - Enterprise AI Solution and Cognitive Engineering Team

AI Engineer
Minneapolis, MN | Sep 2020 - May 2022

• Customer Dwell Activity Detection and Shrinkage Detection System:
  • Detected product theft in POS areas and monitored customer dwell activities.
  • Developed a customer dwell activity detector on retail surveillance data using object detection and tracking.
  • Implemented shrinkage detection by counting objects and comparing them with POS transaction data.
• FAQ Retrieval System:
  • Developed a query retrieval system for frequently asked customer questions.
  • Utilized SentBert to generate embeddings of questions and customer queries.
  • Deployed the model using Django API framework on Azure App Services.
• Other Projects/POCs:
  • Worked on video emotion recognition and active learning POCs.
  • Tools and Algorithms Used: Python, Tensorflow, OpenCV, Pandas.
  • Cloud Tools: Azure App Services.

University of Minnesota - Artificial Intelligence I

Graduate Teaching Assistant
Minneapolis, MN | Aug 2019 - May 2020

• Managed a class of 120 students, assisted the professor with assignments, grading, and resolving questions.

Chegg Inc.

Data Science Intern
San Francisco, USA | Jun 2019 - Aug 2019

• Improved more than 60% of Chegg Study data by building an end-to-end computer vision pipeline for classifying image data according to quality and source, and ROI cropping.
  • Achieved 90.85% accuracy on the image classification model.
  • Tools and Algorithms Used: Python, PyTorch, OpenCV, Pandas, AWS SageMaker GT, Image Processing, CNN.

Indian Institute of Technology, Indore

R&D Intern
Indore, India | May 2017 - Jul 2017

• Developed a 3-D ear recognition model for person identification.
  • Tools and Algorithms Used: Python, Tensorflow, CNN.

Bhabha Atomic Research Centre

Summer Intern
Mumbai, India | May 2016 - Jul 2016

• Computed disparity maps of scenes and integrated 3D coordinates into real scenes.
  • Reduced depth error by 77.5% (from 4cm to 9mm, max allowable error was 2cm).
  • Tools Used: C++, OpenCV, OpenGL, CUDA.

Education

M.Sc in Computer Science
University of Minnesota, May 2020
Courses: Machine Learning, Advanced Artificial Intelligence, Time Series Analysis, Data Mining, Computer Vision, Big Data Engineering, Non-Linear Optimization

B.Tech in Computer Science & Engineering
National Institute of Technology, Durgapur, May 2018
Courses: Software Engineering, Data Structures, Database Management System, Digital Image Processing

Selected Projects

• Stock Market Strategy Prediction using Reinforcement Learning: Explored RL algorithms for optimizing stock portfolios using model-free algorithms and their DNN variations: Q-Learning and Policy Gradient.
• Precipitation Prediction Project: Applied statistical methods (ARIMA) and deep learning algorithms (LSTM and GRU variants) to predict precipitation on Seattle weather data.
• Roof Classification in Aerial Images: Formulated roof material classification in aerial images as a multi-class segmentation problem, handling challenges like very-high resolution (VHR) and missing annotations with CNN-based architectures and selective loss functions to reduce noisy labels.