Lead Machine Learning Engineer

• May 2021 - Present

HFT System • Led a team to design and implement an end-to-end High Frequency Trading system from scratch deployed in equities, futures and options across India, Brazil and China financial markets • Architected and deployed data processing, data modelling, simulation, model validation, model deployment and market feedback pipelines • Developed ultra-low latency (sub microsec) features to extract meaningful signals for price prediction • Conducted rigorous A/B testing… Show more HFT System • Led a team to design and implement an end-to-end High Frequency Trading system from scratch deployed in equities, futures and options across India, Brazil and China financial markets • Architected and deployed data processing, data modelling, simulation, model validation, model deployment and market feedback pipelines • Developed ultra-low latency (sub microsec) features to extract meaningful signals for price prediction • Conducted rigorous A/B testing and experimentation to optimize model performance • Scaled up the strategy to 1000s of stocks with a daily capital of 100 Crore Rs • Achieved average monthly PNL of 2 Crore Rs across asset classes with lower than 1% loss making days Order-Fill Service • Led the strategic vision and implementation of an order execution system for external clients needing fills sized 10 lakh to 1 Crore Rs daily with minimal market impact • Leveraged existing HFT system to achieve 99% fill ratio across clients over a period of 1 year • Setup a system to receive continuous requirements from clients throughout the day, a dashboard to visualise the day’s fills and daily reports to be sent to clients • Onboarded 10s of clients each after an initial POC/test period for the evaluation of the service Asset Clustering • Built a clustering based system to group stocks based on recent price movements and market activity • Developed novel indices based on clusters to be fed into features for data modelling • Achieved improvement of 5% in R-square scores and 10% PNL increase in equity segment Analysis and Visualisation Library • Led the efforts to design and launch a python library which provides numerous tools to analyse and visualise daily stock data, data modelling and analyse the firm’s daily trading • Conducted training sessions for new engineers providing guidance to improve adoption of the library

Senior Machine Learning Engineer

• May 2019 - May 2021

Working as a trader and researcher, developing high frequency trading strategies in Equities, Futures and Options Latency Optimisation • Led the efforts to measure latency of key areas of the trading system and identify high-impact areas • Re-implemented parts of the design and code to decrease packet-to-packet latency by 25% leading to an improvement in PNL of 10% across the board Outlier Detection • Designed and implemented a gaussian density based outlier… Show more Working as a trader and researcher, developing high frequency trading strategies in Equities, Futures and Options Latency Optimisation • Led the efforts to measure latency of key areas of the trading system and identify high-impact areas • Re-implemented parts of the design and code to decrease packet-to-packet latency by 25% leading to an improvement in PNL of 10% across the board Outlier Detection • Designed and implemented a gaussian density based outlier detection system to identify stocks undergoing a change in trading patterns • Dynamically modified the exposure to outlier stocks in daily trading leading to 10% reduction in losses

Machine Learning Engineer

• Jun 2016 - May 2019

Received Star Performer Award - 2018/19 for direct impact on company revenue through effective use of models to improve conversion rates and sign new clients Query Intent Understanding • Designed, developed and deployed an end-to-end prediction system for entity recognition from short search queries by users across e-commerce verticals • Built and deployed DL sequence models and improved relevance of search results by boosting/filtering results based on entities… Show more Received Star Performer Award - 2018/19 for direct impact on company revenue through effective use of models to improve conversion rates and sign new clients Query Intent Understanding • Designed, developed and deployed an end-to-end prediction system for entity recognition from short search queries by users across e-commerce verticals • Built and deployed DL sequence models and improved relevance of search results by boosting/filtering results based on entities identified • Improved conversion rates by 10% and NDCG scores by 25% across Fashion, Home and Living verticals Product Ranking • Developed ranking models for search results based on LambdaMART model • Surveyed existing literature to explore various Learning to Rank methodologies • Generated ranking scores for training data based on metrics defined using session data Semantically Similar Queries • Developed an agglomerative clustering algorithm to identify semantically similar queries • Designed latent query and user similarity features based on bipartite graph walk leveraging Neo4j Graph database • Reduced zero-result queries by 10% leading to improvement in conversion rates and user experience Recommendations • Designed matrix factorisation and collaborative-filtering based recommendation system to power Bought-also-Bought (BAB), Viewed-also-Viewed (VAV) widgets • Improved overall clickthrough rates by 3-5% for fashion vertical Spelling Correction • Designed a spelling correction system to provide suggestions “as-you-type” for short user queries • Trained and deployed a sequence-to-sequence model based on RNNs • Achieved an accuracy of 97% leading to reduction in zero-result queries by 10% Customer Solutions • Led a team to analyse and identify issues in newly onboarded clients’ existing systems • Developed solutions across search and recommendations to deliver a 5-10% improvement in conversion rates within a month Show less Received Star Performer Award - 2018/19 for direct impact on company revenue through effective use of models to improve conversion rates and sign new clients Query Intent Understanding • Designed, developed and deployed an end-to-end prediction system for entity recognition from short search queries by users across e-commerce verticals • Built and deployed DL sequence models and improved relevance of search results by boosting/filtering results based on entities… Show more Received Star Performer Award - 2018/19 for direct impact on company revenue through effective use of models to improve conversion rates and sign new clients Query Intent Understanding • Designed, developed and deployed an end-to-end prediction system for entity recognition from short search queries by users across e-commerce verticals • Built and deployed DL sequence models and improved relevance of search results by boosting/filtering results based on entities identified • Improved conversion rates by 10% and NDCG scores by 25% across Fashion, Home and Living verticals Product Ranking • Developed ranking models for search results based on LambdaMART model • Surveyed existing literature to explore various Learning to Rank methodologies • Generated ranking scores for training data based on metrics defined using session data Semantically Similar Queries • Developed an agglomerative clustering algorithm to identify semantically similar queries • Designed latent query and user similarity features based on bipartite graph walk leveraging Neo4j Graph database • Reduced zero-result queries by 10% leading to improvement in conversion rates and user experience Recommendations • Designed matrix factorisation and collaborative-filtering based recommendation system to power Bought-also-Bought (BAB), Viewed-also-Viewed (VAV) widgets • Improved overall clickthrough rates by 3-5% for fashion vertical Spelling Correction • Designed a spelling correction system to provide suggestions “as-you-type” for short user queries • Trained and deployed a sequence-to-sequence model based on RNNs • Achieved an accuracy of 97% leading to reduction in zero-result queries by 10% Customer Solutions • Led a team to analyse and identify issues in newly onboarded clients’ existing systems • Developed solutions across search and recommendations to deliver a 5-10% improvement in conversion rates within a month Show less

Quantitative Analyst

• Nov 2014 - May 2016

• Developed daily rebalancing strategies by identifying predictive signals in GDP, IP, Political and Lobbying activity, Earnings Revisions, Industry and Cross-Industry Momentum, Brand Visibility datasets • Designed and implemented tests to flag outliers in incoming data across 100s of data vendors as part of the Data Integrity framework; framework used by Portfolio Managers to identify potential issues before finalizing daily positions • Worked on Goldman’s proprietary language Slang to… Show more • Developed daily rebalancing strategies by identifying predictive signals in GDP, IP, Political and Lobbying activity, Earnings Revisions, Industry and Cross-Industry Momentum, Brand Visibility datasets • Designed and implemented tests to flag outliers in incoming data across 100s of data vendors as part of the Data Integrity framework; framework used by Portfolio Managers to identify potential issues before finalizing daily positions • Worked on Goldman’s proprietary language Slang to develop models and perform research Show less • Developed daily rebalancing strategies by identifying predictive signals in GDP, IP, Political and Lobbying activity, Earnings Revisions, Industry and Cross-Industry Momentum, Brand Visibility datasets • Designed and implemented tests to flag outliers in incoming data across 100s of data vendors as part of the Data Integrity framework; framework used by Portfolio Managers to identify potential issues before finalizing daily positions • Worked on Goldman’s proprietary language Slang to… Show more • Developed daily rebalancing strategies by identifying predictive signals in GDP, IP, Political and Lobbying activity, Earnings Revisions, Industry and Cross-Industry Momentum, Brand Visibility datasets • Designed and implemented tests to flag outliers in incoming data across 100s of data vendors as part of the Data Integrity framework; framework used by Portfolio Managers to identify potential issues before finalizing daily positions • Worked on Goldman’s proprietary language Slang to develop models and perform research Show less

Quantitative Researcher

• Jul 2013 - Apr 2014

• Devised robust predictive alphas to exploit reversion and momentum signals in Price-Volume data of global equity markets • Developed low-turnover, event-driven models based on News, Analyst, Fundamental, Options, Twitter datasets • Optimised alphas by performing hedging (market/industry/sector neutralization) and statistical operations on daily positions • Surveyed financial literature to exploit market inefficiencies; developed models to generate significant excess returns at high… Show more • Devised robust predictive alphas to exploit reversion and momentum signals in Price-Volume data of global equity markets • Developed low-turnover, event-driven models based on News, Analyst, Fundamental, Options, Twitter datasets • Optimised alphas by performing hedging (market/industry/sector neutralization) and statistical operations on daily positions • Surveyed financial literature to exploit market inefficiencies; developed models to generate significant excess returns at high Sharpe ratio and low correlation with existing alphas Show less • Devised robust predictive alphas to exploit reversion and momentum signals in Price-Volume data of global equity markets • Developed low-turnover, event-driven models based on News, Analyst, Fundamental, Options, Twitter datasets • Optimised alphas by performing hedging (market/industry/sector neutralization) and statistical operations on daily positions • Surveyed financial literature to exploit market inefficiencies; developed models to generate significant excess returns at high… Show more • Devised robust predictive alphas to exploit reversion and momentum signals in Price-Volume data of global equity markets • Developed low-turnover, event-driven models based on News, Analyst, Fundamental, Options, Twitter datasets • Optimised alphas by performing hedging (market/industry/sector neutralization) and statistical operations on daily positions • Surveyed financial literature to exploit market inefficiencies; developed models to generate significant excess returns at high Sharpe ratio and low correlation with existing alphas Show less

Research Intern

• May 2012 - Jul 2012

Research Intern

• May 2011 - Jul 2011

• Devised an agglomerative clustering algorithm optimizing modularity based on Simulated Annealing technique • Developed high performance parallel implementations for shared-memory multi-core Intel x86 platforms using OpenMP and for the CrayXMT supercomputing architecture • Worked on large-scale real-world networks including Twitter data and US Road Networks • Devised an agglomerative clustering algorithm optimizing modularity based on Simulated Annealing technique • Developed high performance parallel implementations for shared-memory multi-core Intel x86 platforms using OpenMP and for the CrayXMT supercomputing architecture • Worked on large-scale real-world networks including Twitter data and US Road Networks