I am Debasish Das. I did my BTech from IIT Kharagpur and PhD from Northwestern University.

An old version of my webpage and details on my graduate work is maintained at Northwestern EECS.

My resume - longer CV upon request: Updated Nov 10th, 2017

Email me

My current interests include scaling distributed convex optimization, developing machine learning algorithms
for batch/streaming workflows and serving the prediction output at scale. I am working on mobile advertising,
marketing analytics and network security related projects at Verizon Labs, Palo Alto

Spark Summit East 2017

• Hybrid row/column store support in LuceneDAO
• OLAP queries support using sketches in LuceneDAO
• Performance comparisons of Spark + Parquet with LuceneDAO
• Audience demand forecasting using k-NN algorithm

Spark Summit Europe 2016
Trapezium API introduced

• akka-http based configuration driven Spark-as-a-Service
• Spark and Lucene integration (LuceneDAO) for real-time distributed document extraction
• Near-RealTime Lookalike modeling using extracted documents
• Near-RealTime Linear Discrimnant Analysis using extracted documents

Hadoop Summit Tokyo 2016
Trapezium introduced

• Spark Based Lambda Architecture for batch and streaming workflows
• DDOS detection on SNMP and Netflow datasets using Trapezium
• Time Series Anomaly detection using statistical, autoregressive and k-NN models

Spark Meetup 2015
Topics covered:

• Similarity Computation
• Memory based recommendation and comparisons with Implicit Matrix Factorization
• Segmentation using Implicit Matrix Factorization with constraints

Large scale row similarity calculation with user defined kernels (Cosine, RBF)

Matrix Factorization recommendAll for serving cached recommendations

Spark Summit 2014: Quadratic Minimization in Matrix Factorization
Collaborator: Santanu Das

Distributed Matrix Factorization with Quadratic Minimization

Large rank matrix factorization with Nonlinear (Loglikelihood) loss and constraints

Quadratic Programming using ADMM based Proximal Algorithm in Breeze

NonlinearMinimizer using Projection and Proximal Operators in Breeze

JNI library and Java/Scala drivers for Second Order Cone Programming