Amazon Web Services & ipushpull: Revolutionising Financial Data Workflows

The tech world is abuzz with innovation, and at ipushpull, we're thrilled to be at the epicentre of it. After years of building our data integration platform, end-user application integrations and our own chatbot framework our platform is in the right place at the right time (to steal our own tagline!) to embrace GenAI.


We recently embarked on a pioneering journey, joining forces with Amazon Web Services (AWS). Our partnership aimed at harnessing the capabilities of AWS's new AI service, Bedrock, to develop advanced proof-of-concepts (PoCs). Let's delve deeper into this collaboration and the fascinating outcomes it yielded.


The Spark of Collaboration

The prototyping project marked a six-week intense engagement with AWS's European prototyping team and an esteemed mutual customer. Our primary goal? Implementing Large Language Models (LLMs) to facilitate "Named Entity Recognition" (NER) – an NLP process that identifies and extracts financial data points from unstructured natural language within chat platform messages.

Adding to the excitement, we were among the privileged few granted access to the beta version of Bedrock. This new generative AI service integrates a myriad of LLMs, allowing users to harness their power through a single endpoint. Throughout this engagement, our team collaborated seamlessly with AWS engineers and generative AI experts, culminating in a production-ready PoC. The target? Efficient extraction of financial data points from chat messages.


Beyond Prototyping: Expanding Horizons with Bedrock

Our journey didn't end with the prototyping project. Riding on its success, ipushpull was invited to participate further in Bedrock's beta phase. We saw potential and expanded our horizons to create other generative AI-based workflows, specifically tailored for various facets of the Capital Markets Industry.

The below use cases used few-shot prompting techniques against foundation models to allow rapid development of solutions without the complexity and expense of fine-tuning or other customization.


Dive into the Use-Cases

1. Chat-based Natural-Language Query of Securities Reference Data

    • Objective: Enable users to intuitively query data using natural language, while ensuring data providers' access control and swift responsiveness.

    • The Magic Behind: We incorporated Anthropic’s Claude V2 LLM via AWS Bedrock. This converted NLP to ipp platform queries, ensuring swift and accurate data delivery to the end user via chat.


MicrosoftTeams-image (23) 1

MicrosoftTeams-image (24)

Example natural-language queries using AWS Bedrock


2. Trade Booking from Chat

    • Objective: Seamlessly extract trade confirmations from unstructured chat messages, preparing them for ingestion into OMS/trade booking platforms.

    • Behind the Scenes: The LLM was implemented for dual NLP tasks: Classification (distinguishing messages containing trade data from general chats) and NER (extracting trade attributes). Recognised trades were instantly fed into the ipushpull platform, ready for further omnichannel support.

    • The Powerhouse Model: Anthropic’s Claude V1 provided an optimal blend of real-time extraction and model capability.

3. Axe Ingestion from Chat

    • Objective: Recognise traders' interests in IRS markets without imposing a stringent syntax.

    • The Process: Using generative AI, we enabled the platform to identify a spectrum of tenors in IRS axes. 

    • The Workflow: Much like the earlier use case, we extracted chat entities, enriched them, and updated the ipushpull data grid blotter.



Axes request sent via a Symphony chatbot


Working with AWS has been a real privilege from start to finish. We want to thank the AWS team for their support in this project!


“It was a pleasure to work with the ipushpull team during the AWS prototyping engagement, leveraging Amazon Bedrock to build an end-to-end production-ready GenAI solution in under five weeks. I am proud to see what the AWS and ipushpull partnership will foster beyond this engagement to help customers, enabling GenAI workloads on AWS.”

Gaurav Kaila, Prototyping Regional Manager, UK & Ireland, Amazon Web Services


In Conclusion

Our collaboration with AWS and our foray into the realms of Bedrock's AI capabilities has been nothing short of exhilarating. We're more committed than ever to push boundaries, explore new frontiers, and create solutions that resonate with the evolving needs of the Capital Markets Industry.

To use AI, you need data - now ipushpull becomes the launchpad for both. Building on the ipushpull data integration platform, client application integrations and our own chatbot framework meant that LLMs worked seamlessly.

Stay tuned as we continue our journey of innovation and transformation!

Contact us today for more information on how you could benefit from ipushpull

Stay informed with our newsletter