Nilaksh Bajpai

Nilaksh Bajpai

Bengaluru, Karnataka, India
6K followers 500+ connections

Join to view profile

About

At the helm of Flipkart's Top of Funnel technology organization, my leadership extends…

Activity

Join now to see all activity

Experience

  • Flipkart Graphic

    Flipkart

    Bengaluru, Karnataka, India

  • -

    Bengaluru, Karnataka, India

  • -

  • -

  • -

    Bengaluru Area, India

  • -

    Bengaluru Area, India

  • -

    Bengaluru, Karnataka, India

  • -

    Gurgaon, India

  • -

    Gurgaon, India

  • -

    Gurgaon, India

  • -

    Gurgaon, India

  • -

    Noida, Uttar Pradesh, India

  • -

    Lexington, Kentucky, United States

  • -

    Cincinnati, Ohio, United States

  • -

    Mumbai, Maharashtra, India

  • -

    Mumbai, Maharashtra, India

Education

Licenses & Certifications

Publications

Patents

  • SYSTEM AND METHOD FOR DYNAMIC QUERY SUBSTITUTION

    Issued IN 201841003404

    A system for dynamic substitution of a user query adapted for use in an online fashion platform is provided. The system includes a query retrieval module configured to retrieve the user query from the online fashion platform. The user query is provided by a user and the user query has one or more query attributes. The system also includes a session data module configured to store session data collected over a plurality of sessions on the online fashion platform. The system further includes an…

    A system for dynamic substitution of a user query adapted for use in an online fashion platform is provided. The system includes a query retrieval module configured to retrieve the user query from the online fashion platform. The user query is provided by a user and the user query has one or more query attributes. The system also includes a session data module configured to store session data collected over a plurality of sessions on the online fashion platform. The system further includes an analytics module coupled with the session data module and the query retrieval module and is configured to analyze the user query to identify and extract a query object and one or more query attributes. The analytics module is configured to compute a popularity score for each query attribute. The popularity score is a function of the number of times an article related to the product attribute was searched. The analytics module is further configured to compute an affinity score between each query attribute and related product attributes. The affinity score between two related product attributes is a function of approximate substitution of one product attribute with another. In addition, the analytics module is configured to generate a weighted entity-affinity relationship graph (EARG) based on the popularity scores and the affinity scores. Furthermore, the system includes a query substitution module configured to generate a plurality of substitute queries sorted in order of closeness to the user query using EARG.

    Other inventors
    See patent

Projects

  • qwerty jaro winkler

    Other creators
    See project

  • Semantic Search

    - Present

    Semantic Search, code named Charles, is semantic understanding of users query using AI.

    At times, query terms from users can be different from the underlying taxonomy and search may end up showing suboptimal results. Since most of these keywords reside in the high-intent long tail of search queries, they can result in severe disappointment. To avoid this Semantic Search tries to infer the user’s intent from the words in their search query.

    To that end, we built a an…

    Semantic Search, code named Charles, is semantic understanding of users query using AI.

    At times, query terms from users can be different from the underlying taxonomy and search may end up showing suboptimal results. Since most of these keywords reside in the high-intent long tail of search queries, they can result in severe disappointment. To avoid this Semantic Search tries to infer the user’s intent from the words in their search query.

    To that end, we built a an intelligent query understanding module which sits in a query pipeline built as part of this project. It does the following:
    1/ Sanity: Spell check, Stemming
    2/ Query correction: Synonymy, lexical corrections, colloquial understanding etc
    3/ Query understanding: Named entity recognition at scale, classification
    4/ Disambiguator: Entity disambiguation
    5/ Query execution: Forming a Solr query and executing it (taking care of ranking using LTR)
    6/ Query substitution: Substituting query terms using nearness

    Tech Stack: Spring boot, Akka actor model, Spark mllib, Noisy channel, refined soundex, QWERTY edit distance, SolrTextTagger, Arango DB, Solr cloud 6.x

    Other creators
    See project

  • Intelligent payment router

    -

    Intelligent payment router uses a classification model to pick the best payment gateway for a Credit or Debit card transaction. It is trained using historical success rates and predicts the best match for a BIN. It improved the overall success rate by 10 percentage points and has helped during payment gateway outages by intelligently routing to available gateways.
    It uses the bandit algorithm to ensure a good distribution between explore and exploit.

    Tech stack: Logistic regression…

    Intelligent payment router uses a classification model to pick the best payment gateway for a Credit or Debit card transaction. It is trained using historical success rates and predicts the best match for a BIN. It improved the overall success rate by 10 percentage points and has helped during payment gateway outages by intelligently routing to available gateways.
    It uses the bandit algorithm to ensure a good distribution between explore and exploit.

    Tech stack: Logistic regression model in Python, Quartz scheduler in Java to read the model and load in memory

    Other creators

  • Payment Plan Service

    -

    Payment Plan Service is responsible for creating a payment plan and takes care of all types of payments in Myntra. It is a highly scalable, secure and multi tenant system. It uses a Finite State Machine to model a payment process. A customer while placing an order can choose a mix of payment instruments like Credit Card, Debit Card, Gift Cards, Wallets or CashBack Points, PPS uses this information and business (or legal) rules to apportion money. This plan is then executed and by charging the…

    Payment Plan Service is responsible for creating a payment plan and takes care of all types of payments in Myntra. It is a highly scalable, secure and multi tenant system. It uses a Finite State Machine to model a payment process. A customer while placing an order can choose a mix of payment instruments like Credit Card, Debit Card, Gift Cards, Wallets or CashBack Points, PPS uses this information and business (or legal) rules to apportion money. This plan is then executed and by charging the payment instruments selected by the customer. For instance, if the customer chooses to pay by Credit Card and Cashback points then payment gateway is called and apportioned money is charged.
    On cancellation/return of an item, it would ensure to refund the instrument mix that has been used by the customer while placing the order.
    PPS is a a centralized service which takes care of payments in the complete lifecycle of an order. Since it is a mix of debit/credits on various financial instruments, the operations had to be atomic and idempotent.

    Tech stack: JAX-RS (cxf implementation), Statefulj for Finite State Machine

    Other creators

  • Server side cookie store

    -

    • Server side data store + service to replace cookie based system for tracking ROI
    • SLA 25K+ cookies, TTL 48 hrs
    • Tech stack: Redis + Twemproxy, Nodejs

  • Offline Query classification

    -

    • Create query classification model using Mahout
    • Seed appropriate services with category info based on SEM/SEO traffic
    • Estimated saving of $900K in infra cost
    • Tech stack: Apache Mahout, Hadoop etc

  • LaunchPad Tags

    -

    • Estimated saving of ~$350K
    • Container for pixels and administration
    • SLA of ~60 request per second, 100% availability, less than 100ms
    • Tech stack: Spring boot, thymeleaf, tomcat, mysql

  • Yield Engineering

    -

    • Increase monetization of Shopzilla's CSE sites
    • Improve organic traffic (SEO) on Bizrate.com & Beso.com
    • Bizrate.com at one point was top ranked CSE (60K sessions per day)
    • Other experiments (A/B tests) on improving yield, XML sitemaps, hijax, rich metadata

  • Cloud Migration

    -

    • Migrated Shopzilla’s SOA & Web layer to Cloud
    • Joyent used as the IaaS provider
    • Decisions for data security – what & how to expose. E.g. Migrated data from Oracle to Mysql, moved SOLR behind web based interface, etc.
    • Aimed at saving infra cost

Languages

  • English

    Full professional proficiency

  • Hindi

    Native or bilingual proficiency

Recommendations received

13 people have recommended Nilaksh

Join now to view

More activity by Nilaksh

View Nilaksh’s full profile

  • See who you know in common
  • Get introduced
  • Contact Nilaksh directly
Join to view full profile

Other similar profiles

Explore more posts

Explore collaborative articles

We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.

Explore More

Add new skills with these courses

See all courses