Edwin Thomas Email & Phone Number

Toronto, Ontario, Canada

Ottawa, Ontario, Canada

• Architected and extended solutions for supply chain problems using Graph Machine Learning algorithms. • Proposed DICE an intelligent clustering engine for uniformly partitioning directed graphs.• Collaborated in building a smart recommendation system using Reinforcement Learning to facilitate planning operations in supply-chains with ~30x improved performance.

Ontario, Canada

• Conducted systematic survey, replicated and benchmarked recent approaches to Absent Key-phrase generation and Key-phrase extraction.• Explored KPG model evaluation techniques, conducted overlap analysis and statistical significance testing (Bootstrap & Permutation tests).• Studied and implemented various generative model variants including Variational Autoencoders (VAE), Non-Autoregressive models (NARs), Encoder-Decoder architectures, Prefix-Language Models and Decoder-only architectures.• Proposed a new permutation invariant Transformer Decoder with improved attention mechansisms to improve diversity of key-phrase generation achieving SOTA results on KPG tasks (published in NAACL 2024, LREC-COLING 2024).

Ontario, Canada

𝐒𝐄𝐆𝟒𝟏𝟒𝟓𝐁 𝐑𝐞𝐚𝐥 𝐓𝐢𝐦𝐞 𝐄𝐦𝐛𝐞𝐝𝐝𝐞𝐝 𝐒𝐲𝐬𝐭𝐞𝐦 𝐃𝐞𝐬𝐢𝐠𝐧• Conduct lab sessions, tutorials, proctoring and correction of term exams.

Ontario, Canada

𝐈𝐓𝐈𝟏𝟏𝟐𝟎𝐀 𝐈𝐧𝐭𝐫𝐨𝐝𝐮𝐜𝐭𝐢𝐨𝐧 𝐭𝐨 𝐂𝐨𝐦𝐩𝐮𝐭𝐢𝐧𝐠• Conduct lab tutorials, proctor term exams and grading assignments/projects.

Bengaluru, Karnataka, India

• Developed, trained and deployed large-scaled Object Detection and Image classification models for OCI Vision Service for detection of 2000+ real-world concepts with SOTA accuracies.• Proposed the custom training strategy for OCI Image Analysis capabilities, benchmarked quick mode strategies obtaining ~50% reduction in training times with comparable performances.• Developed and improved OCR (Optical Character Recognition) models for Government ID Machine Readable Zone Recognition. • Implemented Graph Neural Networks for key-value extraction from visually rich documents including Driver License documents (50 US states, UK and Canada). • OCI AI Hackathon Winner, recognized for the exceptional application of Deep Steganography for Invisible Digital Watermarking

Bengaluru, Karnataka, India

• Implemented data collection pipeline using Data Crawler, web content extraction, and post-processing algorithms, and used it to boost Named Entity Recognition model performances by ~20%.• Contributed towards fine-tuning Aspect based Sentiment Analysis models offered under OCI Language Services. • Built a Linux command line tool for labeling NER datasets, increasing labeling efficiency by ~40%.

India

• Developed a responsive web-based books and records management system using Angular 7 and backend using Java SE 8.• Enabled data catalog search feature using Natural Language Processing Algorithms.

Master Of Computer Science, Applied Artificial Intelligence, 9.9/10.00

Bachelor Of Technology - Btech, Computer Engineering, 9.68/10.00

Education record