Platform Architecture

End-to-End Reconciliation Workflow

FopsAI is an AI-native reconciliation and ETL platform — built for performance, governance, and auditability in regulated financial operations.

AI Onboarding Agent
Live

Don't configure these stages manually. Our AI Onboarding Agent walks you through the entire workflow in a single conversation — from raw file upload to live reconciliation. Every decision is explained and auditable.

Learn more

Four Stages. One Platform.

From raw file to matched results — each stage is AI-augmented and connected. The AI Onboarding Agent can guide you through configuring all four in a single conversation.

STAGE 1Live

Data Ingestion & Transform

AI Capabilities:

✓ AI Transform Agent

✓ Semantic column mapping

✓ Natural language rule creation

Integrated ETL:

✓ Multi-channel extraction (file, email, SFTP)

✓ 50+ operators with Polars (Rust) processing

✓ AI data profiling and quality analysis

Onboarding Agent profiles your data, detects column types, identifies securities, and recommends cleaning rules automatically.

STAGE 2Live

Match & Reconcile

AI Capabilities:

✓ AI Match Agent

✓ Semantic match (embedding-based)

✓ Strategy recommendation

No-Code Matching:

✓ 30+ match operators

✓ 6 match strategies (1:1, 1:M, M:M)

✓ Multi-pass (Exact/Fuzzy/Semantic)

Onboarding Agent selects the right strategy and configures match rules based on your data structure.

STAGE 3Live

Break Management

Live Now:

✓ Automatic break detection

✓ Lifecycle & SLA tracking

✓ Reason/resolution taxonomy

✓ Priority escalation & approvals

✓ Evidence capture & audit trail

In Development:

AI Break Research Agent — autonomously investigates break causes by fetching market data and corporate actions. Proposes resolutions with evidence.

STAGE 4Live

Reporting & Monitor

Dashboards:

✓ Daily recon results

✓ Cash proofing

✓ Match history

✓ Break management

Monitoring:

✓ Real-time status

✓ Metrics & alerting

Beyond Batch Processing

Intraday & Near-Real-Time Reconciliation

Don't wait until tomorrow morning to find out about today's breaks. Legacy tools run overnight. FopsAI runs when your data arrives.

File-Arrival Triggers

New files trigger reconciliation runs immediately — no waiting for the nightly batch.

Intraday Re-Reconciliation

As corrections or late trades arrive, FopsAI re-runs matching and updates results automatically.

Continuous Monitoring

SLA tracking, health checks, and priority alerting ensure your team is always ahead of exceptions.

T+1 Ready

Built for the accelerated settlement cycle — process and match intraday to catch issues before settlement cutoff.

Multi-Asset Class

From Liquid Equities to Complex OTC Derivatives

FopsAI handles the full spectrum of financial instruments. The combination of 50+ transform operators, semantic matching, and flexible schema configuration means FopsAI adapts to your data — not the other way around.

Standard Products

Equities & ETFs

ISIN/CUSIP/SEDOL resolution, corporate actions, position reconciliation

Fixed Income

Bond identifiers, accrued interest, coupon date matching

Cash & Treasury

Multi-currency reconciliation, value date matching, bank statement processing

Complex / Non-Standard Products

OTC Derivatives & Swaps

Aggregate matching for netted positions, semantic matching for trade terms, tolerance-based notional comparison

Private Debt & Credit

Flexible schemas for non-standard loan tapes, reference table lookups, amortisation schedule transforms

Fund Administration

NAV reconciliation, multi-custodian position matching, investor-level transaction matching

Built Different

Cloud-Native Architecture

Legacy reconciliation platforms were built for on-premise data centres and retrofitted for cloud. FopsAI was built cloud-native from day one.

Serverless Scaling

AWS ECS Fargate auto-scales compute based on demand. Process 100 rows or 10 million rows with the same infrastructure.

Rust-Powered Data Engine

Polars (built in Rust) processes data 10-100x faster than Python/Pandas or Java/SQL engines. Million-row datasets in seconds.

AI Without Compromise

AWS Bedrock provides LLM inference within your VPC. No data leaves your environment. Provider-agnostic — switch models without code changes.

Infrastructure as Code

Every component is Terraform-managed. Reproducible deployments, version-controlled infrastructure, and disaster recovery by design.

Private Deployment Option

Deploy in your own AWS VPC or on-premise for full data sovereignty. Same platform, same features, your infrastructure.

Enterprise & AI-Native Architecture

Live

LLM Service (Provider-Agnostic)

✓ AWS Bedrock (Anthropic Claude)

✓ Azure OpenAI, Groq & open-source models

✓ Centralized gateway with automatic fallback

✓ Privacy-first: no data used for model training

✓ Cost metering & usage tracking

Live

Security & Compliance

✓ Multi-tenant isolation

✓ JWT authentication with RBAC

✓ Immutable audit trails (WORM-ready)

✓ Maker-Checker workflows

✓ TLS 1.3 / AES-256 encryption

Live

Scalability & Performance

✓ ECS Fargate serverless containers

✓ Polars (Rust) vectorized processing

✓ Dedicated Celery worker pools

✓ Redis caching & session management

✓ Table partitioning for high-volume data

Launching 2026

Self-Service Agent Studio

Build your own reconciliation agents with a visual canvas. Drag-and-drop nodes — LLM, tools, conditions, approvals — to create custom workflows for your specific operational needs. No code required.

Why FopsAI Over Legacy Platforms?

Legacy PlatformsFopsAI
Setup4-12 weeks with vendor professional servicesAI Onboarding Agent configures through conversation
Data PrepRequires separate ETL tools and scriptsIntegrated ETL with 50+ operators — no separate tools
MatchingRules only (brittle, breaks on data changes)3-Pillar: Exact + Fuzzy + Semantic AI (resilient)
AI ApproachPredictive / pattern-based, opaqueReasoning with Glass Box — every decision explained
AdaptabilityManual rule updates, support ticketsAI agents adapt, flag changes for approval
Audit TrailPartial, often opaqueEvery AI decision explained with reasoning (Glass Box)
ProcessingJava/SQL engines, nightly batchPolars (Rust) — intraday + file-arrival triggers
Pricing$200K+ annual licence, multi-year contractsFrom $2,500/month, no lock-in
CloudRetrofitted for cloud, or managed hostingCloud-native (AWS), serverless, infrastructure-as-code
AI PrivacyVaries, often opaqueData never trains models. VPC-contained inference.