Shankar Venkitachalam
I am a Staff Machine Learning Engineer at Adobe, where I lead projects on Adobe GenStudio, focusing on the use of Large Language Models (LLMs) for brand-aligned marketing content creation at scale for enterprises. I have developed deep learning models for text and image understanding that power Adobe's Firefly Content Tagging Services.
Previously, I earned my graduate degree in Computer Science from UMass Amherst, where I worked with Andrew McCallum and Pedram Rooshenas on Structured Prediction in NLP.
Before UMass, I worked as a Machine Learning Engineer in the Multimedia Research Group at Snapdeal, Bangalore, and as a software developer at Adobe India. I hold a Bachelor's in Computer Science from the National Institute of Technology Calicut.
Work Experience
|
Staff Machine Learning Engineer Lead the development of Adobe GenStudio from ideation to launch, utilizing generative AI models for scalable, on-brand, personalized marketing content creation. Finetuned and aligned LLMs on marketing content from brands for creating brand-adherent content. Optimized models for creativity, lexical diversity, personalization and domain knowledge. Also developed models and pipelines for validation of content adherence to brand style and guidelines. |
|
|
Senior Machine Learning Engineer Created models for content understanding and tagging of text and images, offered as services via APIs, as well as integrated into Adobe products such as AEM. Developed NLP models for Keyphrase extraction, Named Entity recognition and text image classification, and CV models for image classification and color extraction. |
|
|
Machine Learning Engineer Devised a novel conditioning strategy for generation of images conditioned on given semantic attributes using unconditional GANs. Published a paper in AICC workshop at CVPR 2021. Engineered model and pipeline for semi-supervised identification and summarization of marketing segments and critical event detection in online web sessions. Filed two patents and a workshop paper. |
|
|
Machine Learning Engineer Intern Developed anomaly detection and cause analysis models for Creative Cloud marketing and customer journey data, leading to a patent. |
|
Senior Software Engineer, Machine Learning Built solutions for automatic validation and standardization of new listings. This included text classifiers for category prediction, as well as deep learning models for background extraction and harmful content detection. Also implemented pipelines for feature extraction, clustering and nearest neighbor computation of catalog images, to be used for image-based search and product similarity. |
|
|
Member of Technical Staff II Played a key role in building Adobe Learning Manager (formerly Adobe Captivate Prime), from the ground up. Designed and built the backend system and APIs for processing and storage of course content. Integrated third party services for transcoding videos and converting documents to HTML. Also created the administrator role UI. |
|
Software Engineer, R&D Implemented features in the software layer for enterprise ethernet and optical networking switches and routers. |
Education
|
University of Massachusetts Amherst |
|
National Institute of Technology Calicut |
Publications
|
Directional GAN: A novel conditioning strategy for generative networks |
|
CrEOS: Identifying critical events in online sessions |
Patents
|
Generating digital content consistent with context-specific guidelines utilizing prompt augmentation and model tuning |
|
Generative model-assisted content generation and interactive content editing |
|
Using Shapley values to evaluate prompt generation parameters |
|
Generating weighted contextual themes to guide unsupervised keyphrase relevance models |
|
Quantifying and improving the performance of computation-based classifiers |
|
Generating Image Metadata Using A Compact Color Space |
|
Machine learning models applied to interaction data for facilitating modifications to online environments |