Divya Punj, Natural Language Processing (NLP) Developer in Plano, TX, United States
Divya Punj

Natural Language Processing (NLP) Developer in Plano, TX, United States

Member since February 12, 2019
Divya Punj is a senior data scientist and developer with 12 years of experience in predictive analytics covering a wide range of domains from capital market risk management to customer segmentation in eCommerce, both at the individual contributor and leader/manager level.
Divya is now available for hire



  • Python 3 6 years
  • R 6 years
  • Natural Language Processing (NLP) 6 years
  • Scikit-learn 6 years
  • NumPy 6 years
  • Pandas 6 years
  • Agile 5 years
  • H2O Deep Learning Platform 2 years


Plano, TX, United States



Preferred Environment

Mac, Python

The most amazing...

...problem I've solved for, at the height of credit crisis, was "probability of default," used in calculation of CDS.


  • Head of Data Science

    2018 - 2019
    Aviall, A Boeing Company
    • Designed, implemented, and maintained a smart quoting system, which dynamically scores the probability of an RFQ to be converted into an actual sale. The system was prototyped in R and Python and then fully developed on H2O over a Hadoop cluster. The implementation has been done by deploying a Java executable.
    • Designed and implemented the most probable value of various aviation parts and consumable. The model takes in all the historical as well as static data of the airlines' partners and predicts the most probable price at which the supplies can be procured within the required timeframe. The model requires prediction of requirements based on seasonality, fleet mix, and airline operation and uses algorithms including random forest, GBM, custom ANN, and more. The major achievement was in implementation, where I built out a model to pick the right model to ensure that there is no failure due to old models being used or a model going wrong due to changes in the nature of the data. All of it is automated.
    • Built prediction of prices of airplane spare/replacement parts. Currently, there is no way to find out the fair market value of these products.
    • Used NLP and LDA analysis to analyze client requirements and co-joint analysis.
    • Developed smart quoting, resulting in an efficiency of USD $15 million in addition to revenue per month at its lowest performance.
    • Created the MPV project, resulting in an added profit of USD $8+ million in Phase 1 of the implementation.
    • Scaled both the projects to all of Boeing's subsidiaries for supply chain efficiency.
    • Developed the FMV project, bringing in USD $32 million more in revenue value.
    Technologies: Python, H2O, R
  • Head of Analytics, Special Projects Office of the CEO

    2016 - 2018
    Sears Holding Corporation
    • Designed the order delivery algorithm for all the deliveries to be made by the SWY relay app. Dynamic creation of multiple routes for deliveries was optimized to minimize delivery expenses. The solution had real travel time incorporated using Google Maps API.
    • Created scenario analysis for any new business initiative to understand the zone of profitable operations—a typical analysis would consist of 50+ billion scenarios.
    • Created a technician routing algorithm for Sears' home services division. This required conceptualization of the entire route planning into parallel space-time continuum which converges/overlaps for certain conditions and then diverges again.
    • Created a predictive analysis algorithm which would predict the products that a customer would be looking to buy on SYW Relay, thereby reducing the time taken to order by members. Predictive analysis on SYW Relay has resulted in extremely high continuity levels of the customer cohort, i.e., more than two orders per month from every customer.
    • Used NLP and LDA to understand the feedback of customers and isolate the problem areas. This entailed topic modeling and sentiment analysis both at the same time.
    • Saved $10 million by creating the delivery routing algorithm for the SYW Relay business.
    • Saved $22 million by creating the technician routing algorithm.
    • I had designed a chatbot while I was working at Sears. This chatbot was able to understand the intent of the user and provide a suitable response. For the same, we had created an intent map based on past interactions with the customers. We had a yearly saving of 2 million dollars for this project.
    Technologies: Python, H2O, R, NLP, ANN
  • Head of Analytics

    2015 - 2016
    • Optimized CPC to achieve the lowest possible values.
    • Analyzed the click-through rate (CTR) to increase the conversion of visitors into buyers.
    • Optimized for better conversion rate.
    • Fine-tuned Facebook campaigns to increase conversions by 250%.
    • Increased CTR by 300%.
    • Reduced the online CPC to 30% of its original value.
    • Increased customer engagement by 200%.
    Technologies: R, Python, Machine Learning, Digital Marketing Optimization
  • Manager Analytics

    2014 - 2015
    • Analyzed CTR to increase conversion. Created customer segmentation analytics resulting in an increase in conversions by up to 100% and at least 50% in sub-categories of the fashion division.
    • Created customized predictions of related products for each segment of customers, resulting in higher website traction.
    • Optimized the website for better conversion rate.
    • Used cluster analysis to segment customers for better prediction of behavior.
    • Create a new data analytics process for internal use to identify process lags or process lapses.
    • Increased the CTR by 20%.
    • Introduced new product features which increased the gifting category sales by 200% and musical instruments by 70%.
    Technologies: Python, R, Omiature
  • Lead Analyst

    2012 - 2014
    • Used machine learning (LDA and NLP) to understand the sentiment of each event on pricing.
    • Created Excel-based models to test for the analytics output of the system.
    • Create cross-functional MIS reports which would help various departments leverage each other’s capabilities.
    • Generated an additional revenue of $4 million.
    • Created two new product lines.
    Technologies: R, Python, VBA
  • Senior Research Manager

    2011 - 2012
    • Created Voice of the HNW Client: Customer Interaction Analysis, a strategic study on understanding the requirements and expectations of high-net-worth clients from their asset management advisors and firms.
    • Wrote Channel Volumes: Channel Preference Analysis (Segmentation), an engagement designed to help banking strategists to understand the emerging trends in the consumer banking channels. Identified and built capabilities to succeed in dynamically evolving banking channel preferences for the consumers.
    • Published Consumer Financial Monitor, which tracks quarterly sentiments and current financial status of financial consumer segments across the world.
    • Managed research projects including preparing timelines, scoping, creating, and delivering strategic research.
    • Managed and advised over 150 heads of strategy at banks and asset management organizations across North America and Europe.
    Technologies: SPSS, R, Python
  • Solutions Manager

    2007 - 2008
    Calypso Technologies
    • Created requisite analytical models to support the forecasting, pricing, and risk management needs of the clients for the EMEA region.
    • Provided pricing expertise on exotic derivatives products in interest rates.
    • Provided middle office risk management expertise.
    Technologies: Python, VBA
  • Finance Engineer

    2006 - 2007
    Pyxis Technologies
    • Launched structured European long-dated options for a leading brokerage firm.
    • Worked on an external consulting project for a leading brokerage firm to launch their new range of structured derivatives products. It involved time series extension of Black Scholes implied volatility surface, using Garch (1,1) and ARIMA models. Worked on applying the Heston model as well.
    Technologies: Matlab, Python
  • Software Engineer

    2005 - 2006
    Tavant Technologies
    • Resolved problems proactively and retroactively in the application to enable 24X7 functionality.
    • Helped gather requirements and design the workflow process for the client's mortgage product SNAP, which dealt with customer acquisition and retention. Here, I gained an insight into all programming skills.
    Technologies: Java, J2EE, SQL


  • NLP and LDA Analysis to Understand Customer Feedback (Other amazing things)

    Typically, organizations seek customer feedback using surveys. A lot of potential data is lost using this method because most people either simply don't respond or try to respond and find that the things they would like to communicate are not possible given the survey's prescriptive framework.

    This issue is mitigated by analyzing customer contact center data or freeform text feedback from customers to glean not only the information that a customer would seek via surveys but also much richer data that would not have been covered there. A company using this method will save potentially millions of dollars replacing their customer surveys with this sort of method.


  • Languages

    Python 3, R
  • Libraries/APIs

    Pandas, NumPy, Scikit-learn
  • Paradigms

  • Platforms

    H2O Deep Learning Platform
  • Other

    Natural Language Processing (NLP), Discriminant Analysis (LDA), Agile Data Science


  • MBA in Finance and Marketing
    2009 - 2011
    FMS Delhi - Delhi
  • Bachelor of Technology degree in Engineering Physics
    2001 - 2005
    IIT Bombay - Bombay

To view more profiles

Join Toptal
Share it with others