Data Scientist with 7+ years of experince in building data intensive machine learning applications in diverse industries. Passionate machine learning professional and data-driven analyst with the ability to apply ML techniques and leverage algorithms to solve real-world business problems.
WORK EXPERIENCE
Senior Data Scientist
05/2018 to Present
Blue Yonder India Pvt. Ltd. Bangalore
1) Dynamic Safety Stock Project:
Objective : Its a great challenge for retailer to show correct inventory on e-commerce website for online sales due to various reasons that lead to inconsisteny between true inventory and system-shown inventory. The objective of this project is predict the inconsistency in the inventory so that retailer can show correct inventory that results in high sales and low pick-declines.
I have handled this project from scratch starting with data analysis, feature engineering, model architecture, model evaluation, report generation that helps clients understand our predictions and interacting with clients explaining the model impact along with answering their questions.
Used Google BigQuery to host and analyse the data. Apache Beam Data Flow jobs for preprocessing. Tensorflow Keras model architecture that predicts skellam distribution(explored) parameters.
Written a Custom loss function for tensorflow probability Skellam library for my model.
Used Tensorflow Extended Pipeline (TFX)that creates a DAG of various components in the pipeline. Kubeflow orchestrater to execute the TFX pipeline till a 2000+ page report generation.
Engaged myself with various Client interactions during sales presentations, after go-live issues, etc. This model brought down pick-declines by 15% while increasing sales by 10% for an ongoing client.
2) Probabilistic Availability :
Objective : The objective of this project is also in similar lines with Safety Stock project. But, this solution gives a probability of an order being serviced at a given store/location.
I have come up with absolutely important features that send strong signal for the classifier model. Worked on building light-weight features in Beam Pipeline that makes the model near real-time. Built end-to-end TFX Pipeline right from data pull from BigQuery till the model evalution and report. Client Interactions for pre-sales and custom implementations.
3) Demad Forecasting:
This was one of my first projects at BY and I started with classic ML algorithms such as RANDOM FOREST, ARIMA algorithms
Scikit-learn based libraries for modelling and evaluation purposes.
Then moved on DNN and RNN based models. This is when I introduced tfRecords for storing data.
Experimented with Deep Learnig models for Distributed Training with Parameter Server Strategy.
Data Sceince Engineer
09/2017 to 05/2018
Genpact India Pvt Bangalore
Holistic Promotion Optimizer Project:
Objective: The objective of this project is increase the viewership of a TV channel by optimally sequencing the program advertisements.
My role was to analyse the data and contribute to the engineering side of the project.
I contributed as Engineer in implementing the back-end system for this project.
Supported the post go-live activities uncluding analysing the impact of the model performance along with issue fixing.
Algorithm Engineer
05/2015 to 09/2017
Nuvizz Software Solutions Bangalore
1) Route Optimization for Package movement Without Time Constraints
Objective : Given a set of package occupancy data and vehicle capacity data, the algorithm should minimize the number of vehicles used and minimize total distance travelled.
Worked on Implementing Genetic Algorithm to come-up with an sub-optimal route as initial solution. Used Hierarchical Clustering for grouping the geo-locations before the Genetic Algorithm execution. K-Nearest Neighbor algorithm and few other swapping techniques to arrive at optimal route.
Worked on incorporating Google Maps travel time estimation into this project while optimizing the routes.
2) Route Optimization for package movement with Time Constraints : -
Objective : This is a different algorithm from previous project as delivery time window obligations need to be met while optimally delivering the packages.
Implemented operators such as Simple Insertions, Relocation Operator and 2_Opt operator that swap the destinations within the boundaries of time constraints.
SKILLS
Python, SQL, Matplotlib, Numpy, Pandas, ScikitLearn, Docker
Deep Learning Frameworks: Tensorflow, Tensorflow extended, Google Storage, Google BigQuery, Apache Beam, Kubeflow, Keras, Tensorfow
Decision Trees, Random Forest, ARIMA, Genetic Algorithms, Clustering Algorithms,Prophet, DNN,
: -
CNN, RNN
EDUCATION
Sri Sathya Sai Institute of Higher Learning
Master’s
M.Tech in Computer Science Prashanthi Nilayam, India 06/2013 to 04/2015
Sri Sathya Sai Institute of Higher Learning
Master’s
M.Sc. in Mathematics Prashanthi Nilayam, India 06/2011 to 04/2013
Bengaluru / Bangalore (KA)
, India - IST (Indian Standard Time)
8 Years
English ( Fluent ) , Telugu ( Fluent ) , Hindi ( Intermediate ) ,
: 586
Login with Google
