Bellevue, Washington, United States
Experienced software engineer with 7 years in building scalable and secure systems. I currently work on the anti-scraping team at Meta, where I help design and scale rate-limiting infrastructure for all Meta products —handling over a trillion requests daily. Previously, I worked on Facebook Attribution in the Ads org and contributed to the Portal/Messenger assistant, gaining expertise in large-scale distributed systems and backend infrastructure. I hold a Master’s in Computer Science from Virginia Tech, specializing in NLP and text summarization.
Engineering the core blob storage infrastructure at Meta to guarantee durability of data at global scale. Building large-scale distributed systems and data pipelines to monitor storage health, prevent data loss, and ensure constant availability.
I am a software engineer on the anti-scraping team at Meta. My team owns the rate limiting infrastructure for Facebook, Instagram, and Oculus. I also own the request logging platform that helps determine rate limiting thresholds and develop machine learning models to identify bots. I’ve previously also worked on the Facebook Attribution product that helped advertisers measure the effectiveness of their ads.
Built and scaled the security systems that protect Facebook, Instagram, Oculus, and WhatsApp from bots, automated attacks, and unauthorized data scraping. Created large-scale ML infrastructure that collects signals to train models for bot detection.
Worked on NLP and Dialog Systems for Facebook Messenger • Developed a neural coreference resolution system for Facebook Messenger's "M" chatbot • Created a data annotation tool for identifying named entities and linking them as co-referent • Integrated the neural coreference resolution system into a hierarchical sieve-based classifier that allowed for the addition of rules to handle edge cases • Pushed the neural network model to production
I completed by Bachelor Thesis at Technical University, Munich in Biological Natural Language Processing. Along with two other students, I built a Natural Language Processing Framework (nalaf) for named entity recognition and relation extraction from biomedical texts and journals. nalaf is a general-purpose module-based and easy-to-use framework for common text mining tasks. The framework allows training, annotating and predicting using pluggable components. The framework was initially developed for natural language mentions of mutations and relation extraction of transcription factors to genes and gene products and therefore has some focus on the biomedical domain. The current development of the framework however is towards generalizing all parts and making nalaf a generic NLP tool.
I worked with the Cloud Engineering and PaaS team which spearheaded the hybrid cloud journey at PayPal. I worked on PaaS Infrastructure Manager (PiMan), an application to create Mesos-based clusters in a cloud agnostic fashion. PiMan automates provisioning, service discovery, scaling, upgrade, monitoring, alerting and remediation for these clusters in an immutable fashion. PiMan was built on Java and MongoDB, relying on Terraform, Packer and Consul for provisioning, packing images and service discovery respectively. Some of the highlights from my work here are: • Designed and developed a Chaos Monkey framework to inject failures into and test resiliency of PiMan provisioned infrastructure. The different kinds of failures included shutdown, destroy and process termination. The termination schedule could be randomized to mimic real world scenarios. • Implemented MDC logging for PiMan for distinguishing interleaved log outputs from parallel threads. • Setup ELK stack for real time information mining and log visualization I also developed proxC, an application to manage project lifecycle in Google Cloud Platform. The application was self-service where users could perform CRUD operations on GCP projects. I built the application ground-up with SSO integration and well defined authorisation roles. proxC adopted a 3-tier architecture where the API layer was written using the Dropwizard framework (Java 8), the UI layer using Angular 4 and the database as MongoDB.
I interned with the Platform-as-a-Service (PaaS) team at PayPal where I worked on containerization of PayPal applications using Docker. The highlights of my work here include: • Automated the deployment of applications of different stacks (Node.JS and Java) on the developer's computers or virtual machines • Demonstrated a proof-of-concept where multiple applications of different stacks can be run on the same machine, improving resource efficiency • Enhanced the functionality of a ReST service to promote PayPal microservices from QA to Production by including support for Docker-based applications
Graduate Teaching Assistant for CS3214 - Computer Systems, a course that introduces design approaches in operating systems, inter-process communication, multithreading, and network protocols as relevant to modern web applications. My responsibilities include: • Hold office hours where students can get their questions clarified • Conduct tutorials to introduce assignments and projects • Grading assignments, projects, and exams • Setup auto-grading infrastructure for projects and assignments