Ujan Sengupta Email & Phone Number

Seattle, Wa, Us

Amazon RufusMy team focuses on using Generative AI, traditional lexical/semantic search, and high-performance distributed computing to create a shopping assistant that can help customers across a wide spectrum of shopping missions. In my current role, I help lead the technical architecture, implementation, and operational excellence of a variety of work streams, which include:- Enabling internal Amazon businesses and organizations to provide runtime evidence for the underlying LLM to generate helpful shopping recommendations.- Reducing latency of the LLM responses by employing various online and offline caching strategies.- Enabling product managers and scientists to fine-tune the LLM for different shopping missions and customer use-cases to ensure accurate conversational responses.- Creating an experimentation framework and platform to help the organization rapidly iterate and learn from decisions.- Creating entirely new CXes in the Rufus layer to augment the baseline shopping experience on the Amazon mobile app.

Seattle, Wa, Us

Amazon Halo- I was a tech lead in the Content Platform and Discovery space where I led the creation of Halo's search backend systems and its primary content management software, including the ingestion pipeline, the storage, and the distribution channels.Prime Video• I led the architecture and implementation efforts to build Prime Video's primary media metadata matching platform. Key features include: o Multiple matching systems that provide high precision (96%) and recall (85%) for duplicate title detection in the world's largest streaming video catalog (~6 million titles and counting). o A highly available (99.99%) storage system with low latency (p99: ~150 ms) for result storage. o A translation layer built entirely on server-less technologies that converts raw match results into linkage suggestions to be reviewed by humans. o A centralized user interface to help PV Content Operations take action on the generated suggestions to eliminate duplicates from the catalog. Since launch, the system has helped resolve over 10k+ customer-facing duplicates in the PV catalog, directly contributing to an overall reduction in customer refunds for incorrect purchases and improving overall CX.• Played a critical role in the design and implementation of core components of the Digital Video Supply Chain for the Prime Video Catalog, including building the central entity and relationship management system for the entire catalog which regularly handled operations on 100+ million nodes in a distributed graphDB.• Provided active design and development support for PV-wide initiatives like launching Live Events, extending TVOD title offerings to 260 global territories, launching live linear channels on PV. etc.

Armonk, New York, Ny, Us

- Developed a proof-of-concept cognitive application using IBM Watson Technologies - Native iOS (Swift 2.2)- Leveraged the Tone Analyzer, Personality Insights and AlchemyLanguage APIs to analyze chat conversations in real-time. The analysis provided the conversation tone of individual users and real-time buzzwords that allowed for a contextual tagging of the entire conversation.- Developed a POC web application to demonstrate the Trade-Off Analytics API (from IBM Watson) to facilitate Sales Territory Optimization and Planning.- Helped develop and maintain an internal IBM tool that helps Sales Managers organize their sellers and territory planning effectively and interactively.

Amsterdam, Noord-Holland, Nl

- Developed a prototype for a medical health monitoring application using the cross-platform mobile framework Xamarin, using C# to develop both Android and iOS applications.- The functional prototype will now be used to develop the next phase of the Mobile CareGiver application which will enable a shared code-base between the Android and iOS versions of the application- Also developed the first documented real-time waveform rendering on Xamarin and successfully integrated it with the rest of the application (which was made using the Xamarin.Forms API).

Wachtberg, De

- Research intern in the Cognitive Systems division.- Designed and simulated a parsing algorithm for messages sent and received by unmanned surveillance vehicles.- Used the Robot Operating System (ROS) framework extensively to develop the algorithm.- Used Python to create virtual drivers to help automate test data sets for the same.- Did research in loop closure detection algorithms for autonomous bots.

Gb

- Worked with the Risk Advisory Service (Mobile Application Security) team in the Advisory LOS- Assisted in drafting a bid for advising on Bank of India's application security infrastructure

Montreal, Quebec, Ca

- Summer intern in Donetsk, Ukraine- Was responsible for teaching students enrolled in a summer camp program- Subject matter covered topics ranging from conversational English proficiency to basic social sciences and world cultures

Master Of Science (M.S.), Computer Engineering

Bachelor Of Technology (B.Tech.), Electronics And Communication Engineering

Aissce, Science