Telekonnectors / DSNL

High-performance Software

DSNL is in the enterprise telephony conferencing business. Their technology was built on opensource telephony software and they were not satisfied with its performance. Clay Labs designed and developed a new conferencing engine from scratch with performance as the focus. As a part of this, Clay Labs built a memory efficient SIP and RTP protocol stack. The new system exceeded expectations and was capable of handling 8000 simultaneous phone calls on a mid-range desktop class machine. Each phone call involves transferring 100 packets per second, resulting in a total of 800,000 packets per second.

Clay Labs also built a clustering system, so that a cluster of conferencing servers behave as one large logical conferencing server. This clustering system has been tested to scale to 100,000 simultaneous phone calls.

Happyfox Chat

Scalable Distributed System

Happyfox Chat is a web based live-chat service for customer support. The load on the service can increase dramatically even when signing up just a few customers. This is because, customers who have a large number of visitors on their website will impact the chat service too. Clay Labs designed and developed a highly distributed chat service that can scale to 100s of servers.

Search Engine

Ticketgoose is an online bus-ticket booking system. They collect data from various sources and provide an unified interface for finding and booking bus tickets. Clay Labs designed and developed the search user-interface and the search backend.

The user-interface includes innovative features like visual time-line, and mini seat-maps. The search backend required the creation of a custom caching server that can balance two conflicting constraints - freshness of results and speed of results. The search backend was also specifically optimized for presenting seat-availability information for all search results on the page.


Intelligent Web Crawler

For ContractIQ, Clay Labs designed and developed an intelligent web crawler that can accept a company’s website URL, crawl that website, and extract the management team list. The software will enumerate the names and the titles of the team. This project involved Natural Language Processing, Probabilistic Modeling, and Heuristic Design.