Ishaan Kumar

AI Researcher

Toronto, Ontario, Canada3 yrs 11 mos experience

AI EnabledAI ML Practitioner

Key Highlights

Expert in deep learning and audio synthesis technologies.
Published research on efficient GNN techniques for recommendations.
Proven experience in managing end-to-end ML pipelines.

Stackforce AI infers this person is a Deep Learning and AI specialist with a focus on SaaS applications.

Contact

ishaan@descript.com LinkedIn

Skills

Core Skills

Deep LearningText-to-speech SynthesisRecommender SystemsGraph Neural NetworksConvolutional Neural Networks (cnn)MicroservicesDocker

Other Skills

Artificial Intelligence (AI)BashComputer ScienceComputer VisionData ScienceDeep Neural Networks (DNN)Docker ProductsDocker SwarmElasticsearchEngineering ResearchExperimental ResearchFlaskGenerative AIGitGitHub

About

Deep Learning @ Descript | Ex-Mila | IIT Kanpur

Experience

3 yrs 11 mos

Total Experience

1 yr 11 mos

Average Tenure

Current Experience

Descript

Audio Researcher

Sep 2021 – Present · 4 yrs 9 mos · Toronto, Ontario, Canada · Remote

Media Understanding: Setup up evals to measure quality of visual descriptions produced by VLMs. Used these evals to produce token-efficiency curves; which helped balance costs and quality. This work enabled agentic video-editing workflows to work with visual context.
Zero-Shot Speech Cloning: Developed advanced deep learning models capable of zero-shot voice and roomtone cloning, enabling highly accurate audio synthesis in diverse environments.
End-to-End Machine Learning Pipeline Management: Led all stages of model development, including comprehensive data collection, meticulous data processing, extensive model experimentation, inference optimization. Successfully integrated the final model into the backend platform.
Cutting-Edge Audio Technology: Focused on pushing the boundaries of TTS and audio imputation technologies, thereby reducing technology costs and improving user experience within Descript.

Generative AIText-to-Speech SynthesisDeep LearningEngineering ResearchExperimental ResearchPyTorch

Huawei

Associate Researcher

Aug 2020 – Sep 2021 · 1 yr 1 mo · Montreal, Quebec, Canada · Remote

Graph Neural Networks for Recommendation Systems: Designed and implemented cutting-edge Graph Neural Network (GNN) techniques to enhance Huawei's App Store recommendation systems.
Published Research: Published a peer-reviewed SIGIR 2022 publication titled "EFLEC: Efficient Feature-LEakage Correction in GNN-based Recommendation Systems." Link: https://dl.acm.org/doi/10.1145/3477495.3531770. Also contributed to other research projects in the team.

Recommender SystemsGraph Neural NetworksTensorFlow

Afx medical

2 roles

Machine learning developer

Jan 2020 – Aug 2020 · 7 mos

ML workflow implementation
Training and analysis of ML models
Back-end architecture development

Super-resolutionConvolutional Neural Networks (CNN)

Research Intern

May 2019 – Jan 2020 · 8 mos

Developed and validated deep neural-network methods for MRI super-resolution

Mila - quebec artificial intelligence institute

Masters Candidate

Jan 2019 – Aug 2020 · 1 yr 7 mos · Montreal, Canada Area

Juniper networks

Software Engineer 3

Jul 2016 – Nov 2018 · 2 yrs 4 mos · Bengaluru Area, India · On-site

ML Framework/Library Owner: Developed and maintained a modular, extensible ML framework for network fault detection and prediction using sensor telemetry. Evaluated various ML and deep learning algorithms to optimize accuracy and performance.
API Server Micro-Service Lead: Managed the Flask-based micro-service, providing full RESTCONF API support for the YANG-based input data model, ensuring robust and scalable integration.
Build Pipeline & CLI Utility Lead: Owned and streamlined the build pipeline and CLI utility, enhancing development efficiency and deployment consistency.
Node Virtualization: Assessed Docker container runtime for node virtualization, contributing to improved resource utilization and system performance.
Jsnapy Development: Collaborated with v1 adopters to release v1.1, incorporating new features and addressing critical bugs for enhanced functionality.

Microsoft

Software Engineer Intern

May 2015 – Jul 2015 · 2 mos · Hyderabad, Telangana, India

Summer intern in the Visio product team.
Worked closely with product managers and engineering team to prototype different solutions for Visio content distribution & standardization.
Developed a Provider-Hosted SharePoint marketplace application that facilitated content discovery & installation by aggregating content from third-party providers.
Developed a Visio add-in to search and install content directly from the marketplace.

Docker SwarmDockerPythonMicroservices

Purplle.com

Software Engineer Intern

May 2014 – Jul 2014 · 2 mos · Mumbai, Maharashtra, India

Summer intern responsible for writing the product search application and product recommendation system.
Wrote a wrapper library over Elasticsearch PHP API to streamline application development.
Developed a product search application using Elasticsearch. The new product application increased result accuracy.
Developed a hybrid recommendation system for serving product recommendations. The system was developed using Elasticsearch. This helped reduced costs by eliminating the dependency on third party services.