Aug 2023 – Jan 2024 · 5 mos · Remote

• Working on adding contextual sponsored content to chatbots

Mar 2021 – Apr 2023 · 2 yrs 1 mo · Bengaluru, Karnataka, India

• As part of AI team, managing high-scale operations (more than billion predictions/day) across ML model building, maintenance, retraining, A/B testing and recommendation systems
• Resolved high priority issues like ML model offline-online gap, high latency challenges, data leakage and feature drift
• Developed first version of probability calibration for click/conversion model
• Led and mentored a team of engineers (Software and ML) in design/architecture discussions, PR reviews, and performance evaluations, and provided solutions for technical/business challenges

Jun 2020 – Jun 2021 · 1 yr

• Mentoring software developers to transition to Machine learning role. Guiding them to develop industry scale machine learning project

Dec 2016 – Feb 2021 · 4 yrs 2 mos · Bangalore

• Click Prediction module owner, AMO: Programmatic Search Engine Ad buying platform
• Owner of clicks, cost prediction module for search engine advertising
• Responsible for ML model building, deploying, monitoring click model
• Mentored data scientists & interns to develop budget allocation model & click model for sparse data
• Experimented with models like Logistic regression, XG Boost considering tradeoff between accuracy and interpretability of models
• Fashion Image Retrieval using computer vision and deep learning
• Developed a model to retrieve catalog images similar to fashion images in the wild
• (70% top-5 accuracy with 15k images)
• Used structure and rank preserving joint embeddings of target labels and input images
• Experimented with variations of multi-label classification using CNN based techniques, differential learning rates, stochastic gradients with restarts to achieve 92% category classification accuracy
• Content insights using NLP and deep learning
• Generated contextual image embeddings by training model on auxiliary tasks using images and text
• Image embeddings to improve clustering & impression prediction model (R-squared value gain 8%)
• Latent Dirichlet Allocation (LDA) and Named entity recognition (NER) to generate relevant keywords for text documents

Mar 2016 – Sep 2016 · 6 mos · Mumbai Metropolitan Region

• AlphaPay, stands to change how people send and receive money. We solve problems of merchants and users in utmost simple, and secure way using innovations in payment technology.

Aug 2014 – Nov 2015 · 1 yr 3 mos · Mumbai Metropolitan Region

• Conceptualized & launched personalized recommendation engine to predict listings resulting in 20% increase in conversions
• Led a team of designers, data scientists and QAs and worked with Product manager to build end to end ML backend system
• Developed locality suggestion tool based on similar user mining with 80% accuracy
• Improved real estate map by fixing map polygons at locality level

May 2013 – Jun 2014 · 1 yr 1 mo · Gurgoan

• Mobile recharge loan product default analysis for one of the largest telecom operators in India
• Generated relevant stats, recovery graphs for possible reasons of default and low recovery.
• Devised efficient queries and modeling procedure to handle large dataset(10 million user data).
• Developed credit risk scorecard for customer base and rules based on results from modeling techniques like logistic regression, random forest and decision tree.
• Technical consultancy for Credit risk modeling software for one of the oldest software firms in India
• Improved variable discretization technique, to handle variables with outliers and skewed distribution.
• Incorporated variance inflation factor (VIF), condition number test, F-test to detect multicollinearity.
• Built a credit risk model using classification techniques of logistic regression and decision tree.
• Implemented ‘reduced error pruning’ to enhance accuracy of decision tree.
• Comparative analysis of firm’s proprietary software with R, SAS, SPM.
• Conducted Advanced Modeling Training
• Logistic regression and Decision tree: Conducted training for software developers and analysts from
• one of the oldest software firms in India, on model building using logistic regression & decision tree.
• Statistical survival analysis: Trained a group of 65 analysts from leading analytics firm in India on key
• concepts and applications of ‘Statistical survival analysis’.
• Sentiment Analysis and Twitter data extraction
• Analyzed twitter data and used advanced NLP techniques to generate information related to sentiment of a particular brand in market.
• Churn analysis for India’s oldest m-commerce firm
• Used techniques from survival analysis to determine ‘probability of churn’ and ‘time to churn’.
• Model assumptions were satisfied for Cox proportional hazard model and log logistic parametric model.
• Lift chart, Gini index (48%), KS statistic (28 %) & rank ordering was used to determine model
• performance.

Jul 2012 – Feb 2013 · 7 mos · Bangalore

• Helped the firm realize profits by executing trading and risk management strategies in Gasoil.
• Studied oil market supply & demand dynamics to develop predictive intuition of oil prices.
• Discovered patterns in price movements of Gasoil using data from correlated products.
• Studied and applied technical indicators for speculating price movements.

May 2011 – Jul 2011 · 2 mos · Mumbai Metropolitan Region

• Analyzed relation b/w overnight volatility and trading hours’ volatility for S&P 500 stocks
• Discovered cause of incorrect ‘national best bid-offer’ data by correcting functions in database
• Computed correlation b/w ‘price changes in Stocks’ & ‘buyer-seller initiated trade imbalance’ to
• know market impact using S&P 500 stocks using 2010 data