Bringing Generative AI to Capital Markets: OneAlpha's Serverless Financial Document Intelligence Platform

Company Overview

OneAlpha is a fintech company building AI-powered financial analytics and intelligent capital markets solutions. Its work centres on turning financial data and documents into actionable insight for an industry where speed, accuracy, and compliance are non-negotiable. As the business grew, OneAlpha set out to modernise the infrastructure behind its products and to bring Generative AI into production for the first time, aiming to analyse financial documents faster and at lower cost while holding to the security standards expected in capital markets.

Customer Challenge

1. Infrastructure Not Optimised for AI Workloads

• OneAlpha's existing environment was not built for AI processing and could not readily support Generative AI in production.

• The team needed a cloud-native architecture capable of low-latency financial data processing.

2. High Infrastructure Cost and Operational Overhead

• Over-provisioned Amazon EC2 instances ran 24/7 regardless of actual demand, driving high fixed infrastructure costs.

• Manual patching and server management added significant operational overhead.

3. Slow, Manual Financial Document Analysis

• Financial documents were reviewed by hand, taking two to four hours per document.

• This limited how quickly the business could turn documents into usable insight.

4. No Production LLM Capability

• OneAlpha had no large language model capability running in production.

• The company needed Generative AI integrated for document analysis and insight generation, not just experimentation.

5. Security and Compliance Demands of Capital Markets

• Any solution had to maintain the security and compliance standards expected in capital markets.

• This required strict access control, network isolation, and full auditability across every component.

Solution Approach

CloudTry designed and deployed a fully serverless, Bedrock-native Generative AI architecture on AWS, replacing always-on infrastructure with an event-driven pipeline that analyses financial documents in seconds.

1. Generative AI Document Analysis (Amazon Bedrock – Claude Haiku)

• Amazon Bedrock, using Claude Haiku, extracts financial insights from each document via a structured prompt.

• This delivers fast, consistent analysis in place of manual review.

 

2. Event-Driven Serverless Pipeline

• Financial documents are uploaded to Amazon S3, which triggers AWS Lambda functions through S3 Event Notifications.

• Lambda invokes Amazon Bedrock to analyse the document, and the extracted insights are stored in Amazon DynamoDB along with an audit trail.

• OneAlpha's application layer retrieves insights through Amazon API Gateway, backed by Lambda.

 

3. Scalable Serverless AWS Architecture

• Amazon S3 — stores incoming financial documents and triggers processing via event notifications.

• AWS Lambda — orchestrates the pipeline on a pay-per-invocation basis with no idle cost.

• Amazon Bedrock (Claude Haiku) — performs financial insight extraction.

• Amazon DynamoDB — stores extracted insights and the audit trail on on-demand pricing.

• Amazon API Gateway — serves insights to the application layer through a managed, per-request endpoint.

• Amazon VPC — provides network isolation for the pipeline.

• AWS IAM — enforces least-privilege roles scoped to each service.

• Amazon CloudWatch — handles monitoring, logging, and alerting across the full pipeline.

 

4. Structured Prompt Design for Reliable Extraction

• Prompt engineering, refined during the proof-of-concept phase, was used to extract financial insights accurately.

• The structured approach delivered measurably better extraction quality than a generic prompt.

 

5. Security and Compliance by Design

• IAM least-privilege roles were scoped to specific resource ARNs from day one to prevent permission sprawl.

• Amazon VPC isolation and CloudWatch audit logging were applied across all AI components.

 

6. Fully Serverless Operations (No EC2)

• The entire solution runs without EC2, removing infrastructure management overhead.

• Cost follows a pay-per-invocation model that scales directly with document processing volume.

 

The solution runs in a live AWS production environment with fully automated serverless operations and no EC2 instances to manage.

Outcomes & Impact

Transformation Overview (Before vs After)

 

Aspect	Before Generative AI	After Generative AI
Document Analysis	Manual review, 2–4 hours per document	AI-driven analysis in under 60 seconds
Compute Model	Always-on EC2 instances, 24/7	Serverless Lambda, pay-per-invocation
Infrastructure Cost	High and fixed, regardless of demand	An estimated 40–50% lower, scales with usage
Operational Overhead	Manual patching and server management	Zero EC2 to manage, fully automated
LLM Capability	None in production	Amazon Bedrock live in the production pipeline
Security & Monitoring	Limited	IAM least-privilege, VPC isolation, CloudWatch audit logging

 

Results Achieved

1. Document Analysis Reduced from Hours to Under a Minute

• Turnaround dropped from two to four hours of manual review per document to under 60 seconds with Amazon Bedrock.

 

2. Infrastructure Cost Cut by an Estimated 40–50%

• Replacing always-on EC2 instances with serverless Lambda on a pay-per-invocation model removed idle compute spend and lowered infrastructure cost by an estimated 40–50%.

 

3. First LLM Capability Live in Production

• Amazon Bedrock was integrated into OneAlpha's production financial intelligence pipeline - the first LLM capability live in the environment.

 

4. Zero EC2 Infrastructure to Manage

• Post-deployment, OneAlpha runs fully automated serverless operations with no EC2 instances to patch or maintain.

 

5. Strengthened Security Posture

• IAM least-privilege roles, VPC network isolation, and CloudWatch audit logging were applied across all AI components, improving the overall security posture.

 

6. High Customer Satisfaction

• The engagement closed with a customer satisfaction score of 4.75 out of 5.

 

Key Differentiators & Business Impact

This is a production-grade, serverless-first Generative AI implementation built for the demands of capital markets. By eliminating always-on EC2 — the largest source of wasted cloud spend in AI workloads — and replacing it with an event-driven Bedrock pipeline, the solution delivers both speed and a variable cost model that scales with real document volume rather than fixed capacity.

The architecture addresses OneAlpha's core challenges directly: it turns hours of manual document review into sub-minute analysis, brings Generative AI into production for the first time, and removes the operational burden of server management entirely. Investing in structured prompt design during the proof-of-concept phase produced measurably better extraction accuracy than a generic approach, which mattered for the precision financial documents demand.

Security and compliance were treated as design constraints, not afterthoughts. Least-privilege IAM scoped to specific resource ARNs, VPC isolation, and CloudWatch audit logging give OneAlpha an auditable, capital-markets-ready posture across every AI component — alongside the estimated 40–50% reduction in infrastructure cost that made the business case.

 

Technical Capabilities → Business Outcomes

Event-driven Amazon S3 + AWS Lambda pipeline	Fully automated, hands-off document processing
Serverless architecture (no EC2)	An estimated 40–50% lower infrastructure cost; no idle spend
Amazon DynamoDB with audit trail	Traceable, auditable storage of extracted insights
Amazon API Gateway + Lambda	Secure, scalable access to insights for the application layer
AWS IAM least-privilege + Amazon VPC	Secure, scalable access to insights for the application layer
Amazon CloudWatch	Real-time visibility and audit logging across the pipeline

 

Conclusion

Service Partnership

CloudTry helped OneAlpha move from a legacy, server-based setup to a fully serverless, Bedrock-native architecture on AWS. The engagement replaced always-on compute with an event-driven pipeline, integrated Amazon Bedrock for document analysis, and put security and monitoring controls in place across the system — delivering OneAlpha's first production Generative AI capability on a cost model that scales with actual usage.

 

Partner Support Services (Pre- and Post-Implementation)

Before implementation, CloudTry led discovery and scoping, producing a statement of work that included commercial terms and a total cost of ownership analysis comparing OneAlpha's existing EC2-based setup against the proposed serverless architecture. The team designed the end-to-end architecture and ran a proof-of-concept phase focused on structured prompt design, which improved extraction accuracy ahead of production. Engaging OneAlpha's CTO directly in the architecture review — rather than only the technical team — accelerated sign-off and reduced revision cycles.

After deployment, CloudTry handed over a fully automated serverless environment with no EC2 to manage, and delivered Amazon CloudWatch dashboards as a standard part of the engagement, giving OneAlpha's team immediate visibility into pipeline performance from day one. The security controls — IAM least-privilege scoping, VPC isolation, and CloudWatch audit logging — were established as part of the build rather than retrofitted. The engagement ran from March 2026 to April 2026, closing with a customer satisfaction score of 4.75 out of 5.

 

About CloudTry

CloudTry is a cloud and AI services company headquartered in Lucknow and with offices in Noida, and Kathmandu(Nepal), specializing in the design and deployment of cloud-native and generative AI solutions on AWS. As an AWS Advanced Tier Services Partner, CloudTry brings a team certified across the AWS stack, holding the AWS Certified Cloud Practitioner, AWS Certified AI Practitioner, AWS Certified Machine Learning – Specialty, and AWS Certified Solutions Architect credentials at both Associate and Professional levels. The company works with organizations to move from manual, staff-dependent processes to scalable, production-grade AI systems, with a focus on retrieval-augmented generation, conversational AI, and cost-efficient serverless architectures, among other ready-to-deploy AI solutions, fully built on AWS.