Software Developer, Machine Learning
July 2019 - Present- Work as an applied machine learning engineer at Facebook Assistant (AR/VR) in the products of Oculus and Facebook Portal.
- Develop machine learning models, analyze data, process and extract features and integrate them into runtime stack for different product domains in the fields of Information Extraction and Entity Resolution & Ranking.
Graduate Research Assistant
January 2019 - June 2019- As a graduate research assistant, work along with Professor Imannuel Trummer in a joint project of Cornell University and Google Inc to develop a fact checking system using Probabilistic Unsupervised learning.
- Devised an algorithm for language generation using subject-predicate-object clauses which increased the recall by 41.2%
- Publication: Mining an “Anti-Knowledge Base” from Wikipedia Updates with Applications to Fact Checking and Beyond, Proceedings of the VLDB Endowment, Vol. 13, No. 4 [Publication link]
Arcesium, LLC (The D.E. Shaw group)
July 2017 - June 2018- As a full stack developer in Valuation team developed and enhanced various Java, C# and Perl based APIs to support financial modelling of securities belonging to different asset classes, sources (Bloomberg, Reuters and IDC) in an Agile based system.
- Served as the product in-charge of Pricing Addin (Excel COM & Automation plugin) used by D.E. Shaw clients & 4 other teams of the firm. Developed features of user defined functions, price data persistence & upgraded the product to work on a Windows 10 platform.
- Developed a Technical Spike for using Java cogged C++ based Quantlib modules in place of FINCAD services used by JPMorgan Services and Bank of America clients.
- Modified the Model Based Outlier logic to support client requirements for prices of Commodity Forward asset class securities.
- Developed the functionality of using user configured ticker identifiers (RIC, LIN, ISIN) for snapping Reuter feeds using a Perl based API.
Software Developer Intern: Corporate Investment Banking e-Trading
January 2017 - June 2017- Message Console is an e-trading console application displaying all FX trades in real time. Developed and upgraded the existing application to support different views, filters, notifications and alerts to display trade details using Python and Enaml.
- Developed and deployed quickfix-j based client and server models over cloud micro-services (AWS) and added Financial Information eXchange protocol based communication features to existing services. Also, the project included building cucumber test platforms for retrieval of products over AWS.
- In an eTrading environment various order gateways and exchanges communicate with each other using a protocol called Financial Information eXchange (FIX) . Configured an Exchange Simulator and developed an adapter for a new proprietary order gateway which communicates purely in FIX protocol. The project at hand aimed to develop an API that communicates in FIX with the new order gateway which in turn sends messages to various FX exchanges.
Samsung Research Institute Bangalore
May 2016 - July 2016
(Under the guidance of Saidhiraj Amuru)
- Worked in project 5G Communications, simulated coexistence of LTE with Wifi using Network Simulator-3 tool and analysed the difficulties faced and designed methods to overcome the same.
- Implemented the energy detection module using Dynamic spectral duty cycle method in the LTE modules of NS-3 as required for the proposed 5G model by 3GPP (standard organization for implementing 5G technology).
- Analysed and modified the variables affecting the throughput for Wifi with LTE coexistence as required in the proposed 5G model.
Homi Bhabha Centre for Science and Education (TIFR)
May 2015-July 2015(Under the guidance of Prof. G. Nagarjuna and Dr. Shekhar Krishnan)
- Project ChaloBEST is an initiative from the government of India to make transportation data publicly available over the web, SMS, smartphones, and print media using free and open source software. Worked on data wrangling of irregular spreadsheet data of Mumbai city BEST bus services to create a PostgreSQL database from it.
- Used the generated database to construct General Transit Feed Specification, a standard for any transit services platforms like Google Maps, using fuzzing-string matching algorithms like Soundex and python library of fuzzywuzzy.
- Tweaked the fuzzy string matching algorithm of Soundex to incorporate numeric entities.
- Deployed the GTFS standard to generate the a user application over the RESTful API of OpenTripPlanner.
- Implemented a Twitter chat bot conveying bus schedules via user interaction in transliterated Hindi and Marathi language scripts.