Hi, I'm Aishwarya Allada.
A
Self-driven, quick starter, passionate programmer with a curious mind who loves solving a very complex, very challenging real-world problems.
About
Machine Learning Researcher at the University of Waterloo, pursuing MASc with research expertise in Computer Vision and Natural Language processing.
In my academic and professional experience, I have worked on end-to-end analytics projects that involved Data Analysis, Data Engineering, Data Visualization, Data Modeling, Machine Learning Model Deployment, and Analytics Framework Development for solving business problems.
Most recently, I have significantly contributed for Histopathology Image Analysis and NLP for Digital Pathology with Ontario Research Fund – Research Excellence Consortium, as a part of my research. I have experience working with language models for various downstream tasks and experience in feature extraction and data modeling from big data along with knowledge of data visualization tools. I also developed deep learning solutions for solving complex problems and adept at setting up data generation, analysis, and modeling deep learning methodologies.
A keen learner and a hardworking person actively looking for Full-time opportunities in the field of Data Science, for roles like Data Scientist/ Machine Learning Engineer/ AI Engineer.
Please also visit the projects, and academic achievements down in profile I'm proud of. See something you like? I'm right over at aish.allada@gmail.com
Experience
- Currently working on a project in Histopathology Image analysis and NLP for Digital Pathology with ORF-RE Consortium.
- With respect to the pathology reports that are highly unstructured with sophisticated and highly specialized medical terminology, analysis of various contextualized word embeddings and their combinations were demonstrated through the classification task. The results obtained proved that Clinical BioBERT embeddings in combination with TF-IDF feature vectors performed best with the accuracy of 92%.
- Regarding pathology images, a weakly supervised method with attention mechanism is implemented to classify Whole Slide Images having large dimensions using the slide labels, achieving 90% accuracy.
- A method to classify both pathology images with their respective report is proposed, which uses the image and text embeddings and the combination of embeddings proved better in classification resulting 96% accuracy.
- Tools: Python, Keras, Tensorflow, PyTorch, NLTK, Pandas, matplotlib
- Worked on the development of Artificial Intelligence based video conferencing tool. Precisely worked segmentation and video enhancement, and further deployed these features at the browser level using TensorflowJS.
- Tools: Python, Flask, ADD MORE.!!!
- Worked in Lean Six Sigma project, which is an employee management site that will keep a track of the projects and certification done by the employees with Angular 5 and NodeJS. Also, had been a part of Receiving Mastero project where a middleware is created to upload and download files from GCP with NodeJS.
- Tools: Angular5, NodeJS, PostgreSQL
Projects
Created an end-to-end application of a recommendation system along with sentiment analysis of the reviews.
- Based on the users search query, content-based movie recommendation system will recommend similar movies based on the likes and sentiments on the reviews of the searched movie.
- Details of a movie are extracted from TMDB using the API key generated from TMDB.
- Based on the IMDB ID of the movie in the API, reviews are scrapped using the beautifulsoup4 library and sentiment analysis for those reviews are performed using Multinomial Navie Bayes Classifier.
- The application is further deployed on Heroku
An attention-based classification model that aims at classifying the stance of a news and its headline.
Combination of summarization and classification models
- Comparative study by summarizing a huge corpus of a News dataset using seq2seq, pointer generator, BERT, pointer generator with coverage mechanism and further classifying those texts using DNN models.
- Inferred significant improvement in the model’s classification accuracy (94.47%) when a summary of the document is fed rather the entire text..
A slight modification of ADAM optimizer
- The whole experiment shows the contrast of the decay of gradients of original Adam optimizer to the rate of decay of gradients when finite averaging is performed in the AdamE optimizer.
- An innovation for faster decay of earlier gradients is implemented following the same principles of Adam algorithm, so the algorithm remains scalable to high-dimensional machine learning problems.
The research of a novel, efficient and scalable content-based visual media retrieval system.
- Created a comprehensive architecture for both images and videos conjointly, by leveraging the power of Deep Learning to enable modularization and efficiency for various retrieval applications that require scalability as their key requirement.
- Results of two important experiments obtained are - the impact of colours on retrieval tasks and fast comparison techniques that may have a considerable impact on improving retrieval tasks further.
Skills
Languages and Databases
Python
HTML5
Angular5
NodeJS
MySQL
PostgreSQL
Libraries
NumPy
Pandas
OpenCV
scikit-learn
matplotlib
pycaret
Frameworks
Django
Flask
Bootstrap
Keras
TensorFlow
PyTorch
Other
Git
Heroku
Education
Waterloo, Canada
Degree: Master of Applied Science
Specialization: Pattern Analysis and Machine Intelligence
Cumilative GPA: 89.4/100
- Image Processing and Visual Communication
- Methods and Tools for Software Engineering
- Statistical Methods for Data Analytics
- Data and Knowledge Modelling and Analysis
- Introduction to Optimization
- Tools of Intelligent System Design
- Text Analytics
Relevant Courseworks:
SRM Institute of Science and Technology
Chennai, India
Degree: Bachelor of Technology
Specialization: Electronics and Communication Engineering
Cumilative GPA: 98.6/100
- Achieved "Performance Based Scholarship Award" for CGPA.
- Awarded with "Outstanding Overall Performance in Academics and Extracurricular Activities".
- Awarded with "University 2nd Rank" for academic excellence.
AWARDS & RECOGNITION