Machine Learning and Deep Learning Eng

Kavya Pratap Singh

Boston, MA

425-***-**** *****.****@************.*** LinkedIn GitHub

EDUCATION

Northeastern University Fall 2022 – Spring 2024

Masters in Machine Learning GPA: 4.0

Coursework: Machine Learning, Deep Learning & Neural Networks, Computer Vision, NLP, Data Science DIT University, India July 2015 – May 2019

Bachelor of Technology in Computer Science GPA:7.3 TECHNICAL SKILLS

Programming Languages: Python, Experience with SQL and C/C++ Libraries and Frameworks: TensorFlow, PyTorch, OpenCV, Scikit-learn, NumPy, Pandas, Matplotlib, Seaborn, Keras Computer Vision: Convolutional Neural Networks (CNNs), Transfer Learning, Object Detection (SSD, YOLO, Faster R-CNN), Image Segmentation (Mask R-CNN, U-Net), Generative Adversarial Networks (GANs).

Natural Language Processing: Recurrent Neural Networks (RNNs), LSTM, GRU, Transformer Models, Text Generation, LLM Machine Learning: Random Forest, Support Vector Machines (SVM), XGBoost, AdaBoost, K-Means, PCA, Hyperparameter Tuning Data Visualization: Tableau, Power BI

PROFESSIONAL EXPERIENCE

AI Research & Development Intern July 2023 – Present DTonomy Boston, MA

• Integrated, developed and deployed Large Language Models (LLMs) for standalone systems to implement strong and robust data security

• Leveraged transfer learning techniques to efficiently pretrain models on large client datasets, improving accuracy by 15%

• Finetuned LLMs for using state-of-the-art architectures like LLaMA, Falcon, Vicuna improving inference speed by 5%

• Built interactive web-based chatbot interfaces for clients by utilizing Langchain and StreamLit, allowing non-technical users to easily query LLMs through a graphical interface

Senior Engineering Consultant June 2019 – September 2022 Capgemini Kolkata, India

• Developed machine learning models to predict project completion timelines, reducing negotiation cycles by 27.89 days on average

• On employee retention prediction, I engineered features from raw data optimized existing deep learning models, enhancing predictive accuracy by 7%

• To decrease the manual effort of categorizing customer support tickets, I built sorting models using random forest, cutting the time spent by 30%

• Utilizing Pandas and Numpy, object-oriented data processing library was established to prepare datasets for machine learning algorithms

• Delivered training sessions on machine learning to interns and new hires RELEVANT PROJECTS

Lip Reading Computer Vision Model TensorFlow, Deep Neural Networks, LSTM, CNN

• Developed a computer vision lip-reading model and a lip tracking model that transcribes speech from a muted video with 92% accuracy, surpassing previous models by 18%

Predicting Knee Osteoarthritis with Severity Grading Python, TensorFlow, MobileNetV2, Grid Search

• Developed a deep learning algorithm that detects knee osteoarthritis with severity grading with 81% accuracy over 5 classes. Successfully reduced false positives by 27% and false negatives by 19% compared to previous versions. Predicting Age from X Ray Images CNN, TensorFlow, Transfer Learning, Xception, Pandas, Numpy

• Developed an X-ray image age prediction model using Xception CNN architecture that achieved a precise MAE of just 7.3 months, improving on previous approaches by 10%

Modify Image Contents Zero Shot Python, Stable Diffusion, Grounding DINO, Segment Anything Model (SAM)

• Developed an end-to-end workflow for text-based image editing by integrating state-of-the-art AI models - SAM for segmentation, Grounding DINO for object detection, and Stable Diffusion for diffusion-based inpainting Track Football Players and detect Ball Possession Python, Yolov5, ByteTrack

• Developed a real-time football tracking system using ByteTrack and YOLOv5 that improved mAP by 5% and reduced inference time by 20%; added new feature to existing model to detect ball possession in each frame. AI Voice Assistant Python, Whisper, GPT4All

• Implemented a voice assistant in Python using SpeechRecognition, PlaySound, and GPT4ALL; leveraged Whisper for 5% more accurate speech-to-text and Hermes language model for contextual responses.

Weed Detector Python, TensorFlow, Yolov8, YoloNAS

• Developed a real-time 16 class weed detection system using YOLOv8 and YOLONAS, with data augmentation and hyperparameter tuning that improved mAP by over 25% and exceeded Roboflow models by 5% Predicting Flight Delays Python, Deep Learning, Random Forest, Neural Networks

• Developed a neural network model for flight delay prediction that analyzed 500,000 flights, engineered 50+ features, and achieved 89% accuracy, outperforming baselines by 12% and reducing false predictions by 40% Bike Rental Analysis Python, Deep Learning, XGBoost

• Developed XGBoost regression model to forecast bike rental demand from 1M+ records that achieved 82.4% accuracy after data cleaning and feature engineering, improving MAE by 40% and training time by 55% over baselines Housing Rent Predicter Python, Deep Learning, XGBoost

• Implemented Machine Learning models like Random Forest and Linear Regression to predict apartment type with 85% accuracy and estimate rent price within $150 MAE. After hyperparameter tuning and feature engineering improved accuracy by 15% over baseline

Contact this candidate