Skip to content

Case Study / AI / Compliance

ReportAI / ForensicReports / 12 weeks

AI-Powered Due Diligence & Compliance Automation Platform

A serverless AI-powered due diligence platform that automates company research, regulatory data collection, risk analysis, entity matching, and forensic report generation using AWS and generative AI.

  • Serverless Event-driven pipeline. SQS decouples the request lifecycle from research and AI processing.
  • Multi-source Intelligence aggregation. Registries, sanctions, PEP and news sources queried per request.
  • Bedrock AI report generation. Structured due diligence reports from aggregated evidence.
  • AWS
  • Amazon Bedrock
  • Lambda
  • SQS
  • S3
  • DynamoDB
  • API Gateway
  • Node.js
Abstract diagram of an event-driven AWS due diligence pipeline with queue-based workers and AI report generation

Overview

Client Overview

The project required a scalable due diligence platform capable of collecting information about companies from multiple external sources, validating and normalizing the collected data, performing AI-assisted analysis, and producing comprehensive integrity due diligence reports. Traditional manual due diligence requires analysts to search multiple registries, sanctions databases, news sources, corporate records, and other public datasets before manually consolidating the findings into a report. The objective was to turn that fragmented process into a reliable, asynchronous and AI-powered workflow where a user submits a company once and the platform handles the research, processing, analysis and report generation automatically.

Services provided

  • AI & Generative AI
  • Due Diligence Automation
  • AWS Serverless Architecture
  • Data Aggregation
  • Risk Analysis
  • External API Integrations
  • Report Generation
  • Cloud Infrastructure
Client
Confidential — compliance intelligence engagement
Industry
Compliance / Risk Management / Corporate Intelligence
Timeline
12 weeks
Year
2026
Platforms
Web application, REST APIs
Team
2 engineers, 1 cloud architect

Challenge

Business Challenge

The situation before the engagement, in the client's terms.

The platform needed to solve a fundamental problem in corporate due diligence: relevant information is distributed across many sources, while the final report needs to present that information as one coherent, evidence-based assessment.

  • 01

    Fragmented information sources

    Company information can be distributed across corporate registries, sanctions databases, PEP sources, adverse-media providers, news APIs and other third-party services.

  • 02

    Long-running research workflows

    Due diligence requests can require multiple external API calls and substantial AI processing. Running everything inside a synchronous HTTP request could lead to timeouts, poor reliability and a poor user experience.

  • 03

    Manual report preparation

    Raw research data is difficult to consume directly. The system needed to transform structured and unstructured evidence into a consistent, professional due diligence report.

  • 04

    Reliability and failure handling

    External APIs can fail, respond slowly or become unavailable. The architecture needed retries, queue-based processing, status tracking and a dead-letter mechanism.

  • 05

    Traceability of generated reports

    Every report needed a reliable lifecycle: Request → Processing → Research → AI Analysis → Storage → Status → Retrieval.

Objectives

Engagement Objectives

What the build needed to achieve before implementation began.

  • Automate the collection of company and regulatory information

  • Integrate multiple external intelligence and registry APIs

  • Decouple long-running processing from the user-facing API

  • Generate structured due diligence reports using generative AI

  • Store generated reports securely and durably

  • Track report status throughout the complete processing lifecycle

  • Handle failed jobs through retries and a dead-letter queue

  • Build an architecture capable of scaling as report volume increases

  • Provide a foundation for Level 1, Level 2 and Level 3 due diligence workflows

Solution Design

How We Solved It

The approach that addressed each challenge above.

We designed the platform as an event-driven AWS serverless architecture. Instead of keeping the user request open while the entire research and AI pipeline executes, the platform immediately creates a report request, places the work onto Amazon SQS, and processes the request asynchronously through Lambda workers — separating the API layer from the heavy processing layer so the system scales independently.

  • 01

    API-driven report requests

    Amazon API Gateway provides the public API layer. A client submits company information through POST /companyData, a Node.js 20.x Lambda request handler validates it, creates the report record and sends the message to SQS — the client receives an immediate response with the report ID.

  • 02

    Queue-based async processing

    Amazon SQS acts as the processing buffer between the request layer and worker layer, providing decoupling, reliable delivery, retries, traffic buffering, independent scaling and fault tolerance.

  • 03

    Dedicated worker processor

    A separate Lambda worker consumes messages from SQS and orchestrates the research workflow — retrieving the request, calling external data sources, aggregating and normalizing the evidence.

  • 04

    Multi-source intelligence gathering

    The worker integrates News APIs, NewsData.io, MediaStack, Companies House, OpenSanctions, OFAC Trade API, SerpAPI and other registry and intelligence APIs, querying relevant sources selectively by entity and jurisdiction.

  • 05

    AI-powered report generation

    Once research data is normalized, the worker passes evidence to Amazon Bedrock, which converts it into a structured forensic report — executive summary, company identification, business operations, entity matching, ownership, risk assessment, sanctions & PEP, adverse media and appendices.

  • 06

    Durable report storage

    Generated reports are stored in Amazon S3, providing durable access to report artifacts including HTML/PDF outputs and a clean separation between report generation and retrieval.

  • 07

    Metadata & status management

    Amazon DynamoDB maintains report metadata and processing state — report ID, company information, processing status, report location, timestamps and lifecycle data — so the frontend can query status without touching the worker.

  • 08

    Report retrieval APIs

    A separate retrieval flow queries DynamoDB, generates pre-signed S3 URLs where required, and returns the report information or document location — keeping generation independent from consumption.

Capabilities

Key Features

What ReportAI / ForensicReports does in day-to-day use.

  • Automated company research

    A single company request can trigger research across multiple external intelligence and regulatory sources.

  • Multi-source data aggregation

    Information from different APIs is collected, normalized and prepared for analysis within a unified processing workflow.

  • AI-powered due diligence

    Amazon Bedrock analyzes the collected evidence and generates a structured forensic report.

  • Risk assessment

    Findings are organized into risk categories with an overall risk assessment suitable for review.

  • Entity matching

    Evaluates whether information from different sources belongs to the same underlying entity, with confidence surfaced when evidence supports it.

  • Sanctions & PEP screening

    Sanctions and politically exposed person information incorporated through connected intelligence sources.

  • Adverse media analysis

    News and media sources collected and analyzed to identify potentially relevant adverse information.

  • Ownership analysis

    Ownership and related-entity information represented where reliable source data is available.

  • Asynchronous processing

    Long-running workflows execute outside the request-response cycle through SQS and Lambda.

  • Retry & dead-letter handling

    Failed attempts retried automatically; messages exceeding retry limits routed to an SQS Dead Letter Queue.

  • Secure report storage

    Generated reports stored in Amazon S3 rather than inside the application server.

  • Report retrieval

    Completed reports retrieved through dedicated APIs without rerunning the research workflow.

Technology Stack

Technology Stack

  • Amazon Web Services
  • Amazon API Gateway
  • AWS Lambda
  • Amazon SQS
  • Amazon S3
  • Amazon DynamoDB
  • Amazon CloudWatch
  • AWS IAM
  • AWS X-Ray
  • Amazon Bedrock
  • Foundation Models
  • Generative AI
  • AI-assisted Risk Analysis
  • Structured Report Generation
  • Node.js 20.x
  • REST APIs
  • Event-Driven Architecture
  • Asynchronous Processing
  • News APIs
  • NewsData.io
  • MediaStack
  • Companies House
  • OpenSanctions
  • OFAC Trade API
  • SerpAPI
  • Registry & Intelligence APIs

Outcome

Results & Outcome

What changed for the business after launch.

  • Serverless

    Event-driven pipeline

    SQS decouples the request lifecycle from research and AI processing.

  • Multi-source

    Intelligence aggregation

    Registries, sanctions, PEP and news sources queried per request.

  • Bedrock

    AI report generation

    Structured due diligence reports from aggregated evidence.

  • DLQ

    Controlled failure handling

    Retries and a Dead Letter Queue isolate failed jobs for investigation.

Interface

A Closer Look

Screens from the delivered system.

  • Report generation request form
    Report generation — company details, jurisdiction, entity type, report level, data sources and instructions.
  • Report dashboard overview
    Dashboard — report activity, risk distributions, recent reports and AI-generated insights.
  • Reports workspace with searchable records
    Reports workspace — searchable records with level, status, risk score, report ID and actions.

Conclusion

The Result

ReportAI / ForensicReports transforms a fragmented due diligence process into a serverless, event-driven and AI-powered intelligence platform. The architecture combines AWS serverless infrastructure, asynchronous processing, external intelligence APIs and generative AI to create a scalable foundation for automated corporate due diligence — automated research across multiple intelligence and regulatory sources from one company request, Amazon Bedrock converting aggregated evidence into structured reports, SQS and Lambda decoupling user requests from long-running processing, retries and a Dead Letter Queue providing controlled failure handling, and S3 with DynamoDB providing durable reporting and lifecycle tracking.

  • Serverless Event-driven pipeline: SQS decouples the request lifecycle from research and AI processing.
  • Multi-source Intelligence aggregation: Registries, sanctions, PEP and news sources queried per request.
  • Bedrock AI report generation: Structured due diligence reports from aggregated evidence.

More work

Related Projects

Other systems built on similar foundations.

View all projects
  • Abstract diagram of a multi-agent voice routing system with connected conversation nodes

    AI / Voice / Multi-Agent

    VeriVoice

    An intelligent multi-agent voice platform designed to automate business conversations and workflows.

    • React
    • Next.js
    • Node.js
    • AI
    • +1

    View Case Study

  • Abstract diagram of a lead scoring and routing pipeline with branching decision paths

    AI / Automation

    AI Sales Automation

    An automated qualification and outreach engine that scores inbound leads, enriches them and routes each to the right owner within minutes.

    • Python
    • OpenAI
    • n8n
    • PostgreSQL
    • +1

    View Case Study

Next step

Want a System Like This?

Tell us what your business needs to automate and we'll map out a practical build.

Book a Free ConsultationContact Us

hello@thedevrox.comWe reply to every enquiry within one business day.