Download⬇
____________________________________________________________________________
 |
EDUCATION |
|
|
|
UNIVERSITY OF MINNESOTA, Minneapolis, MN |
|
|
|
|
Master
of Science in Business Analytics (Quantitative Methods and Mgmt. Science) (Full-time,
on-campus) |
|
|
|
|
|
|
|
|
|
Z
H College of Engg. & Tech, Aligarh, India |
|
|
|
|
Bachelor
of Tech in Computer Engg. (Full-time,
on-campus) |
|
|
|
|
|
|
|
KEY EXPERIENCE
NETSKOPE – Sr. Staff ML Scientist – Bangalore Dec-2022 – present
AI Labs
- Developed anomaly detection for http network traffic data, using the multivariate time-series model based transformer architecture – to detect the software failures [link].
- Trained Deep learning model with low latency to create a generic guardrail for GenAI applications (LLM based applications) that classifies the GenAI raw traffic
- Feature engineering and model inference pipeline was transitioned in C++ framework to reduce the latency and model-size
- URL embedding model was developed to augment the HTTP traffic classification types
- Developed framework to automate rule generation using the DL model
- Developed RAG pipelines for tasks on Topic Modelling & Question-answering system
- Using LLM such as gemini, claude, & QWEN
- Trained ML model for Image classification for Data Loss prevention use-case and deployed using ONNX C++ transformation
|
Citrix – Principal
Data Scientist - Bangalore |
Jan-2021
- Dec-2022 |
|
|
|
Application
security
| |
|
|
AMAZON – Research
Scientist - Berlin |
Aug-2019
– Dec-2020 |
|
|
Seattle, WA |
Nov-2017 – July-2019 |
|
|
Payments
and Fraud analytics – Amazon Web Services · Designed and trained account risk scoring machine learning model to stop fraudster from using AWS service. This model helped in reducing 10% of manual tasks. · Developed NLP/ML model using word embeddings and classification models for detecting fraudulent websites. This model will help in reducing fraud attacks and creating robust rules for emerging fraud patterns. · Created optimization tool to balance between customer friction, and fraud operational expense – to reduce customer friction with minimum impact on operating expense. ·
Created data-shift detection
scheme – using Kolmogorov-Smirnov test, standardized difference, and KL
Divergence metric – to notify the emerging pattern on a week-to-week basis. Digital
Apps and games – Amazon.com · Developed text matching algorithm for billion possible combination – submitted white paper in Amazon internal conference. This helped in improving the matching accuracy by 20%. ·
Developed the prototype to detect
the possibility of sales spike, based on social media text and historical
sales data - using NLP and deep learning algorithms o Developed recurrent neural network (RNN) to identify sales peak using MxNet ·
Designed and developed prototype
for advertisement keyword generation using tf-idf, page-rank and pos tagging
algorithm. The top result helped in finding new AdWords thus reducing manual work of AdWords
creation. |
|
|
|
Walmart Labs – Statistical
Analyst - Bentonville, AR |
June-2016
– October-2017 |
|
|
·
Developed price impact model, in
Hadoop-R framework. o A
regression-based elasticity/cross-elasticity model at Item-Market area level o This
helped in various price investment strategy to increase revenues in various
categories · Trained fixed effect model to assess impact of feature-promotion, and asses where the promotion should be executed. It helped in improving the campaigns execution by a lift of 5%. · Designed vulnerability management automation, based on machine-learning on Py-Spark framework – selected for Walmart’s internal conference. This helped in reducing 4 manpower of efforts required in ticketing system. ·
Conducted
several ad-hoc, ML analysis for space/shelf optimization, root cause
analysis, cannibalization etc. |
|
|
|
Carlson Analytics Lab - Data
Scientist Intern |
January-2016
– May 2016 |
|
|
·
Developed a
predictive/classification model for a Fortune
100 consumer electronics retail firm- to assign the
customers (100 million records) into the most likely segment, segments
related to technology affinity. The solution was developed using SAS
enterprise-miner, and involved attribute selection from 10,000 attributes,
imputing missing values, and machine learning models. |
|
|
|
Deloitte Consulting Pvt LTD, Bangalore,
India – Consultant (Data Analyst) |
|
|
|
·
Developed reports to measure the
vendor’s performance in terms of time and other item attributes to maintain
lean inventory for Donaldson Inc ·
Developed SQL reports for picking
and shipping process for Donaldson Inc |
|
|
|
ORACLE INDIA PVT LTD,
Bangalore, India - Staff Consultant (Data Analyst) |
|
|
|
·
Developed various supplier
performance reports and web tools for ACT ·
Developed demand forecasting using
Oracle-Demantra for a manufacturing company to schedule supplier orders and
work-order creation to maintain lean inventory. |
|
|
|
TATA CONSULTANCY SERVICE, India and USA |
|
|
|
|
IT Analyst |
|
|
|
|
·
Designed and developed Supply
Chain Management process for Cummins Inc o
Work order (engine manufacturing) plan
creation based on forecasting and sales order to maintain lean inventory
(JIT) o
Sales and Work order pegging(matching)
– to have best fit based on order’s requested ship-date, and scheduled engine
manufacturing completion date. o
Automated the generation of
Barcode reporting at various shipping process to provide failsafe mechanism
such as right order is getting loaded in right truck o
Inventory management automation
based on various signals such backflush, Kanban signals – mini IOT |
||
|
Other fulltime affiliations |
|
|
Senior Software Engineer – Accenture |
Jan-2011, to
Jun-2011 |
|
MBA Drop-out, The State
University of New York at Buffalo |
Aug-2012, to
Dec-2012 |
|
ERP Advisor - Dell International Services |
Feb-2013, to
Nov-2013 |
|
Tools,
Languages, and others |
|
|
|
Python,
R, Py-Spark, Rapid-Miner, JMP, SQL-PL/SQL, Java, C, C++, Hadoop HDFS, Hive,
SAS-enterprise miner |
No comments:
Post a Comment