Page Not Found
Page not found. Your pixels are in another canvas.
Sitemap
A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.
Pages
Page not in menu
This is a page not in th emain menu
Posts
Generative AI - For Beginners
Posted:
Generative AI is transforming various industries by enabling machines to create content like images, music, and text. In this blog post, we'll explore how beginners can easily start using generative AI tools without the need for complex setups or powerful hardware.
No GPU? No Worries!
Posted:
If you’re feeling stuck because you don’t have a GPU, fear not! There are several effective alternatives that can help you run your projects smoothly. In this blog post, we'll explore two excellent options.
Detecting Poison Attacks using Neural Collapse Geometry
Posted:
There are specifically two types of Data Poison Attacks - Untargeted & Targeted. However, detecting untergetted attacks is fairly simple as the overall accuracy of the model decreases. But it becomes a problem when the attack is targeted. Until and unless, the query for that specific label/feature is not triggered, you will never understand that you are hacked! There are methods to identify Targeted Attacks but lets try to figure out if these attacks can be identified / corrected using embeddings’ representation, and thus utilising the Neural Collapse to detect poisoned labels/features.
experience
Graduate Research Assistant Fellow, University of Alberta
Phair Lab | May 2025 - ongoing
- Working on designing efficient, fair de-identification algorithms for clinical text.
- Developing scalable inference pipelines optimized for processing millions of healthcare notes.
- Developing clinical information retention metrics using BERT for de-identified medical text.
Work In Learning Opportunity, Alberta Machine Intelligence Institute
May 2025 - July 2025
- Translated business problems into ML problem statements and solution scopes.
- Conducted research and literature reviews to assess feasibility and risks.
- Supported solution validation, focusing on interpretability and deployment.
University of Alberta
Graduate Teaching Assistant - CMPUT 200 | Sept 2024 - Dec 2024
- Worked as TA for CMPUT 200 - Ethics of Data Science and AI (Fall 2024 & Winter 2025).
- Conducted office hours, labs, grading assignments and exams, solving student queries, etc.
MITACS Globalink Research Internship At University of Calgary
Research Internship | Calgary, Canada | May 2023 - August 2023
- Project: Return on Investment (ROI) of Data Analytics
- Used NLP, Active Learning and requirements dependency extraction to construct a full framework to estimate ROI
- Hosted the application on AWS EC2 and implemented CodePipeline to automate the deployment procedure
- Business Impact: Highlighted possible ROI under several circumstances, potentially helping the business to manage resources and prioritise data-driven plans.
- Tech Stack: Deep Learning, Machine Learning, Data Science, NLP, Cloud Computing (AWS), React JS, Flask
AutomationEdge, India
Machine Learning Internship | India | June 2022 - July 2022
- Conducted data preprocessing, feature engineering, and model evaluation on large text datasets.
- Enhanced text analytics pipelines and improved customer intent recognition models.
- Tech stack: Machine Learning, Deep learning, NLP, Spacy, Textacy
Amazon ML Summer School’22
Amazon | Apprenticeship | July 2022
- Competed with the engineering students in India in programming skills, logical reasoning, mathematics and machine learning assessments and was selected among the top few students for this training session
- Hands-on: Deep Neural Networks, Sequential Models, Unsupervised Learning, Causal Inference & Reinforcement Learnin
projects
Progetto - Project Management Platform
Tool for project management
- Developed a platform to manage all the ongoing projects on a University Campus.
- Designed REST APIs for interactive user experience.
Dense Network Pruning using Neural Collapse under Imbalanced Dataset
Neural Network Pruning Algorithm
- Neural Collapse-Inspired Pruning: Preserves class separability during pruning,in imbalanced datasets.
- Bias Mitigation & Robustness: Reduces bias towards majority classes and tests model robustness under noisy conditions.
- Experimental Results: Demonstrates effective accuracy retention and fairness, especially for minority classes in imbalanced data.
Q&A: System Based on Google Palm LLM and Langchain
LLM-powered Q&A system using Langchain + Google Palm
- Developed an LLM-powered Q&A system (Langchain + Google Palm) within Streamlit UI to reduce workload for e-learning company.
- Spearheaded the implementation of FAISS vector database for fast retrieval of relevant answers, optimizing system performance and user satisfaction.
- This innovative approach enables real-time response and seamless access to knowledge within the e-learning platform.
Vehicle Detection Through Transfer Learning
A Transfer Learning framework for detecting vehicles in aerial images
- Utilised the potential of transfer learning with VGG16 network architecture.
- Attained quicker convergence on vehicle detection dataset.
- Conducted trials on grey and coloured images and obtained model accuracy of 60%.
research
AROhI: an Interactive Toolkit for Estimating ROI of Data Analytics
IEEE 15th Annual Computing and Communication Workshop and Conference (CCWC), Las Vegas, USA 2025
This work details a comprehensive tool that provides conventional and advanced ML approaches for demonstration using requirements dependency extraction and their ROI analysis as use case.
View here
Not What the Doctor Ordered: Surveying LLM-based De-identification and Quantifying Clinical Information Loss
The 2025 Conference on Empirical Methods in Natural Language Processing 2025
This work shows LLM-based de-identification often modifies clinical details. Current metrics to capture such changes are not expert-validated. We propose an expert-validated metric to test over-redaction.
View here
Towards Fair and Efficient De-identification: Quantifying the Efficiency and Generalizability of De-identification Approaches
Accepted to the Findings of The 19th Conference of the European Chapter of the Association for Computational Linguistics 2026 2026
This work shows LLM-based de-identification often modifies clinical details. Current metrics to capture such changes are not expert-validated. We propose an expert-validated metric to test over-redaction.