Husain Zafar

• Build and operate managed Kubernetes at scale on Amazon EKS, with a focus on EKS Fargate.
• Developed an approach for gracefully recycling EKS Fargate customers' nodes to keep them secure, recycling pods one availability zone at a time to avoid service downtime.
• Released Kubernetes 1.22 for EKS Fargate and contributed to region build; improved the Fargate customer experience using Fluent Bit and improved Fargate billing operations.
• Led the effort to automate the EKS Fargate region-build process with zero manual code changes, and to make the Fargate billing system more reliable through increased observability.

• Developed a regional serverless notification service to alert EKS customers about issues with their Kubernetes clusters, eliminating manual effort to do the same.
• Used Step Functions to work at scale; added cross-account, cross-partition authorization to access the service.

• Built a Flask service to wikify CSV files, linking cell labels to their corresponding Wikidata QNodes.
• Formulated an approach for mapping a text label to its corresponding DBpedia URI from column context for the ISWC cell-entity annotation challenge; extracted features using batch SPARQL queries for later approaches.

• Developed a solution to find outliers / insider threats in an organization using peer-group analysis and non-negative matrix factorization; cut runtime several-fold with Apache Spark caching and the Parquet file format.
• Implemented REST APIs for the IGI Trial Offering (a cloud-based solution for instant, on-demand IGI trials); deployed cron jobs to check demo-machine status and schedule trials ahead of time.

• Led development of an image-classification pipeline to detect and classify defects such as roots and cracks in groundwater pipelines.

• Coursework: Algorithms, Artificial Intelligence, Machine Learning, Data Mining, Natural Language Processing, Databases.

• Coursework: Data Structures, Object-Oriented Programming, Discrete Structures.

Smart Music Player: facial emotion & mood recommendation

• Built a music recommendation system that identifies a user's facial emotion to recommend a song from their collection; achieved 97.69% accuracy on the music-classification module. Published at WiSPNET 2017.

Halma AI agent

• Used minimax adversarial search with alpha-beta pruning to play Halma on a 16×16 board, with adaptive search depth and an efficient opening sequence.

OpenAirInterface install, simplified

• Isolated components of OpenAirInterface (5G research software) with Linux network namespaces to enable quick install and build on a single node, removing the need for multiple nodes.

Multiplayer typing-speed tester

• Led a team of three to build a Qt app measuring typing speed, accuracy and WPM in real time, with head-to-head competitions; played by 100+ students at the college tech fest.

• S. Gilda, H. Zafar, C. Soni, K. Waghurdekar. "Smart Music Player Integrating Facial Emotion Recognition and Music Mood Recommendation." WiSPNET, 2017.
• A. Thawani, M. Hu, E. Hu, H. Zafar, N. T. Divvala, A. Singh, E. Qasemi, P. Szekely, J. Pujara. "Entity Linking to Knowledge Graphs to Infer Column Types and Properties." SemTab @ ISWC, 2019.

• Student Volunteer, International Conference on Machine Learning (ICML) · Jun 2019.
• Organized TEDxPICT, Pune · Sep 2017.