Experience time

1-3

As an Applied Scientist at Amazon’s AWS AI/ML division, I specialized in building scalable, production-grade machine learning systems that powered cloud-based AI solutions for global clients. My role involved end-to-end model development—from research and experimentation to deployment—focused on improving the intelligence and performance of services like Amazon SageMaker, recommendation engines, and predictive analytics tools.

I collaborated closely with cross-functional teams to deliver high-impact ML features, optimized models for large-scale inference, and ensured performance through rigorous evaluation and A/B testing. I was also actively involved in initiatives around responsible AI, model fairness, and explainability—making sure our solutions were not only powerful, but trustworthy and ethical.

This experience deepened my ability to bridge advanced research with real-world applications, shaping AI systems that are fast, flexible, and reliable at global scale.

Work Experience

Applied scientist - Machine learning
Amazon August 25, 2023 - Present At Amazon Web Services (AWS), I worked as an Applied Scientist in the AI/ML division, developing scalable machine learning solutions for cloud-based platforms. I contributed to the design and deployment of predictive models for services like Amazon SageMaker and personalized recommendation engines for retail clients. My responsibilities included model experimentation, A/B testing, optimization, and collaborating with product teams to bring AI features from prototype to production. I also worked on initiatives focused on responsible AI and model explainability across AWS tools.

Education

Artificial intelligence
Master of science July 12, 2021 - May 5, 2023 Master of Science in Artificial Intelligence Indian Institute of Technology (IIT) Bombay Specialization: Machine Learning, Deep Learning, Natural Language Processing Graduation Year: 2023

Portfolio

Devansh Kapoor

Scalable Personalized Recommendation Engine for AWS Marketplace

I led the development of a real-time, scalable recommendation system for AWS Marketplace to improve product discovery and personalization for global users. I designed a hybrid model combining collaborative filtering and content-based embeddings, which improved recommendation relevance by 30% based on NDCG and CTR metrics. Using Amazon SageMaker Pipelines, I built end-to-end ML workflows for training, evaluation, and deployment. For inference, I implemented a low-latency, serverless architecture using AWS Lambda and Step Functions, ensuring high availability and scalability. I ran extensive A/B testing across user segments, resulting in a 12% increase in engagement and an 8% boost in conversions. I also incorporated SHAP-based model explainability and worked with responsible AI teams to ensure fairness across different user cohorts.

View Portfolio

Honors & awards

Best Paper Award – NeurIPS / ICML / ACL (as co-author)
October 9, 2024 Honored for co-authoring a research paper on transformer optimization techniques published and presented at a top-tier AI conference.
0 Base on 0 reviews
Clarity in Specification
Requirements were clear, precise, and well-structured.
0
Communication
Responsive, collaborative, and proactive in all interactions
0
Payment Promptness
Payments were timely and handled without issues.
0
Professionalism
Highly organized, respectful, and focused on quality.
0

Reply

Cancel reply