Chandramathi Murugadass

Overview: Graduate Research Assistant working on RNA motif-based RNA prediction as part of the DAIS research group. We predominantly work on DeepTracer a research on protein, DNA, RNA or other biomolecule model prediction an extension of the Nobel prize winning Rosetta model.

• Generate and validate a feasible dataset of RNA motifs. This includes segmenting the 3D motif structures from the whole RNA.

• Use the generated dataset to train an RNA prediction deep learning model.

• Develop a website that allows users to submit prediction requests.

• Teaching Assistant for the Data Structures, Algorithms, and Discrete Mathematics II course under the guidance of Dr. Neelkantan Kartha. Involved in assignment preparation, writing programs for validating submitted code.
• Grader for Autonomous Systems and Vehicles, worked for Dr. Tyler Folsom.

• Company Overview: Zoho WorkDrive is a Document Management and content collaboration software in the cloud. We receive billions of file upload requests and serve users across the globe with strategic data centers in different regions.
• I architected and designed an implementation of lambda architecture a data deployment and batch processing model for real time data, using Apache Hbase as the immutable master of data set and Apache Kafka for stream processing.
• Detect and restrict any REST API requests that are initiated by unruly bots using reinforced machine learning algorithm to detect behavioural anomalies.
• Frequently used metadata, such as the shard of a file, is cached in application servers. Any change in shard information is then propagated through Apache Redis clusters, which are maintained using consistent hashing.
• Zoho WorkDrive is a Document Management and content collaboration software in the cloud. We receive billions of file upload requests and serve users across the globe with strategic data centers in different regions.

• Machine learning is employed for spam detection in files published through WorkDrive. These files may contain spam, phishing, or malicious content, and they are proactively identified using decision trees built through semi-supervised learning.
• Designed and implemented email uploads that allow users to send attachments in an email addressed to a WorkDrive folder, which will be directly uploaded to the folder if the sender has write permission, and if both the sender and the content pass through the spamming filter. The architecture involves POP3 servers that receive the mails and Kafka consumers write the polled attachment to WorkDrive folders.
• Designed, modeled and implemented a data serialization and de-serialization framework that converts the modeled data from data stores and transfers to recipients using protofub and gRPC.
• WorkDrive allows a team admin to track the activity of a user within a team folder or team, designed a data model to fetch this data from a denormalized reverse index built on Elastic Search.

• Designed and implemented a layer that validates user requests to ensure that only authorized users have access to the files stored in WorkDrive. The goal is to ensure that the processing time for this layer is less than 20ms.
• Implemented a distributed queue based on Apache Kafka, which further breaks down into several sink consumers that perform mutually exclusive operations. Each consumer is an independent micro-service and can be independently activated or deactivated. Consumers are used to send mobile or email notifications, compute the storage or trigger other workflows.
• Typically, files are accessible only within a WorkDrive team account. External sharing enables users to publish files by creating unique links, each with multiple configurable properties. These links can be password-protected and configured to expire. The metadata is stored in MySQL but password and other personally identifiable information are encrypted using a user specific key maintained using a Key Management Service.

• Files stored in Zoho File System, can be digitally signed using third party signing service providers such as Adobe Sign and DocuSign.
• The user interface of WorkDrive interacts with the WorkDrive servers through REST APIs. I designed and drafted a specification for the API format based on the JSON API specification. Additionally, I built a framework that converts object model definitions to JSON responses and vice versa.

• Analysis and Implementation of File Compression Zoho File System is the centralized store for all files uploaded to the Zoho cloud, for all products such as Zoho CRM, SalesIQ Projects, etc. In order to optimize the memory used by the applications, the stored files had to be compressed.