Arshad Shaikh Email & Phone Number

United Kingdom

Bengaluru, Karnataka, India

BADRI: Byjus Attentive Datevec Rasch Implementation- Multi-Modal Multi-Objective Transformer-based model trained on over 30B+ interactions.- Uses Knowledge Graph-based embeddings, along with Date aware attention Decay.- Knowledge Retention using Forgetting Curves Prediction Head: "Push Notifications".- Implemented Cached Inference that enabled efficient.

Bengaluru, Karnataka, India

Deep Equation- Implemented SOTA Algorithm for Mathematical Equation Retrieval.- Hierarchical Transformers/LSTMs extracting and encapsulating Equation Embedding. - Improved Contrastive Loss function by Leveraging Knowledge Graphs.

Mumbai, Maharashtra, India

- Rethinking Neural Networks with Benford’s Law [NeurIPS 2021] - Introducing Self-Attention to Target Attentive Graph Neural Networks [AISP 2022]- Knowledge Tracing.- Dataset distribution shift on FormOCR, Improved alignment.

Canada

Search and Recommendation - Efficient Scalable Inference using Quantization and ONNX.- Fitting large models on AWS Lambda- Chatbots RASA

Mumbai, Maharashtra, India

- Image annotation and classification (Microscopic images 10x compound).- Model error analysis and evaluation.- Data Structures ( Created a prototype for a project that is a data structure that loosely and partially mimics our inadequate understanding of memory. Python3 was used with several unit tests on each step. )

Mumbai, Maharashtra, India

- Research Paper published: Deep Learning for Part of Speech (PoS) Tagging: Konkani [springer]- Efficient Serving of the DL Models-Tf (Django)- Automatic Subjective Grading pipeline that scrapes posts from TheCN and scores using ML models. Selenium. - Marathi Poem Generation Using Genetic Algorithms- Surveying Biologically Inspired Algorithms

Master Of Science - Msc, Data Science

Bachelor Of Engineering - Be, Automobile