{"data":{"jobs":{"edges":[{"node":{"frontmatter":{"title":"Software Engineer","company":"Traba","location":"New York City, NY","range":"May 2025 - Present","url":"https://www.traba.work"},"html":"<ul>\n<li>Working in the Agents team, building and scaling Scout, Traba's AI-powered worker vetting agent (<a href=\"https://traba.work/company/engineering/building-scout\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">read about it here</a>).</li>\n</ul>"}},{"node":{"frontmatter":{"title":"Machine Learning Engineer","company":"Two Sigma","location":"London, UK","range":"Aug 2021 - May 2025","url":"https://www.twosigma.com"},"html":"<ul>\n<li>Designed and implemented a Ray-based distributed compute framework that optimizes data locality, resulting in 3x faster Machine Learning research workflows with 50% compute cost reduction.</li>\n<li>Developed and maintained a core library for efficient and reliable model serialization and transfer, streamlining firm-wide model productionization and unlocking regular automated model refits and releases.</li>\n<li>Led the design and implementation of solutions to export 80% of Two Sigma's primary time-series forecasting library's models to production, used by multiple research teams.</li>\n<li>Engineered a one-click system to perform simulations, backtesting and analysis on regularly retrained models.</li>\n<li>Collaborated with ML researchers to develop custom solutions and integrate proprietary ML techniques into our forecasting library.</li>\n<li>Reduced test times for a proprietary distributed compute framework by 85%.</li>\n</ul>"}},{"node":{"frontmatter":{"title":"Software Development Engineer Intern","company":"Amazon","location":"Cambrige, England","range":"Jun - Sep 2020","url":"https://www.amazon.jobs/en/business_categories/alexa"},"html":"<p>I worked on Alexa in the Answer Generation Team as part of a 12-week summer internship, and received a return offer to work at Amazon in the Alexa organisation.</p>\n<ul>\n<li>Worked in <a href=\"https://www.antlr.org/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">ANTLR</a> to implement a strong type system for a custom internal domain-specific language (DSL) used for generating natural language in production systems (Alexa uses the output sentences to answer users’ questions).</li>\n<li>Proposed backwards-compatible syntax change for optional type specification and implemented it by updating parser and lexer rules, internal data structures used for representing the DSL and the parser used to populate these representations.</li>\n<li>Worked in <a href=\"https://www.java.com/en/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Java</a> to implement DSL compiler validations for static type checking, making extensive use of <a href=\"https://github.com/google/guava\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Guava</a>, <a href=\"https://github.com/google/guice\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Guice</a>, <a href=\"https://projectlombok.org/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Lombok</a> and <a href=\"https://junit.org/junit5/\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">JUnit tests</a></li>\n</ul>"}},{"node":{"frontmatter":{"title":"Software Engineer","company":"Scope News Ltd","location":"Cambrige, England","range":"Aug - Dec 2019","url":"https://scopenews.co.uk/get"},"html":"<ul>\n<li>Lead developer of Controllable Abstractive Summariser for News Articles, from data collection till training.</li>\n<li>Worked in PyTorch to adapt MASS by Microsoft which utilizes Transformers for seq2seq deep learning.</li>\n<li>Modified feedforward neural network module to take in extra parameter as an input to control output length</li>\n</ul>"}},{"node":{"frontmatter":{"title":"Data Science Intern","company":"DataProphet","location":"Cape Town, South Africa","range":"Jul - Aug 2019","url":"https://dataprophet.com/"},"html":"<ul>\n<li>Worked in Keras on VAE/CVAE architectures to optimise efficiency of Mercedes-Benz’s C-Class production network.</li>\n<li>Compared and visualised performance of various autoencoder architectures with varied loss functions and metrics.</li>\n<li>Extensively documented Python code using DocStrings in the RestructuredText format.</li>\n</ul>"}},{"node":{"frontmatter":{"title":"Software Engineer Intern","company":"A*STAR I2R","location":"Singapore","range":"Jan - Jun 2018","url":"https://www.a-star.edu.sg/i2r"},"html":"<ul>\n<li>Lead developer of website for labelling LIDAR data with 3D bounding boxes using ThreeJS, Flask, dat.GUI.</li>\n<li>Pre-processed image data (for measurements and de-noising) for prediction algorithm using OpenCV in Python.</li>\n<li>Wrote Bash scripts utilising Robot Operating System and tmux to automate data extraction, conversion and sampling of\n100s of terabytes of LiDAR data and image data collected from LiDAR-mounted car every day for 3 months.</li>\n</ul>"}},{"node":{"frontmatter":{"title":"Administrative Support Assistant","company":"Singapore Armed Forces","location":"Singapore","range":"Jan 2016 - Oct 2017","url":"https://www.mindef.gov.sg/web/portal/mindef/national-service/discover-ns"},"html":"<ul>\n<li>Supported Commander's Office in Signal Institute with various administrative processes as part of <a href=\"https://en.wikipedia.org/wiki/National_service_in_Singapore\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">National Service</a></li>\n<li>Built a fully-automated, data-collection and parsing software using Google Forms and Google Sheets, written in JavaScript\n<ul>\n<li>End-to-end data pipeline; registration process feeds directly into the next part</li>\n<li>Worked 24/7 to capture workout data from over 50 people for over 6 months, automatically tabulate it in a spreadsheet and display statistics</li>\n</ul>\n</li>\n</ul>"}}]}}}