Independent Contractor/AccuNode

AccuNode — Credit Risk Assessment Platform

A production-grade, multi-tenant SaaS platform for ML-powered company default risk prediction.

Overview

AccuNode is a multi-tenant SaaS platform that uses machine learning to predict company default risk from financial ratios. It supports both real-time individual company analysis and large-scale batch processing via CSV/Excel uploads, alongside analytics dashboards for portfolio-level insights.

Built for financial institutions, lenders, and risk teams who need fast, explainable credit risk assessments at scale.

The Problem

Credit risk workflows in most organizations are slow, inconsistent, and hard to scale:

• Spreadsheet-heavy and largely manual • Not designed to handle hundreds of companies at once • Lack ML-powered scoring or customizable models • No multi-tenant access control for teams and organizations

This leads to slow decision-making, inconsistent risk evaluation, and poor visibility across portfolios and sectors.

Target Users

•

Credit & Risk Analysts — need fast, reliable default scoring

•

Lending & Fintech Teams — require scalable batch processing

•

Investment & Research Teams — want portfolio-level insights

•

Platform Admins — manage tenants, organizations, users, and models

How the Platform Works

1.Individual Company Analysis

Users input financial ratios to receive real-time default probability scores and risk categories, powered by ML models trained on annual or quarterly data.

2.Bulk Portfolio Analysis

Users upload CSV or Excel files. The system processes jobs asynchronously, runs batch ML inference, and provides downloadable results with live job status tracking.

3.Analytics Dashboard

Aggregated insights across all analyzed companies: Default rate by sector, Risk category distribution, Default rate vs. market cap, High-risk companies and top performers.

4.Role-Based Access Control

Multi-level roles — super_admin, tenant_admin, org_admin, org_member, and user — control access to data and operations across tenants and organizations.

Architecture & Core Components

  • •API Service (FastAPI) — Async REST APIs handling auth, tenant/org management, predictions, analytics, and job orchestration.
  • •ML Inference Layer — Annual and quarterly default prediction models served via dedicated backend pipelines.
  • •Background Workers — Handle bulk CSV/Excel uploads, feature extraction, and batch inference jobs asynchronously.
  • •Cache Layer — Redis manages job state, caching, and performance optimization across requests.
  • •Relational Data Store — PostgreSQL stores tenants, organizations, users, jobs, predictions, and analytics data.

Tech Stack

Frontend

Next.js 15 (App Router)Next.js 15 (App Router)
React 19 + TypeScriptReact 19 + TypeScript
Tailwind CSS + Radix UITailwind CSS + Radix UI
Zustand + TanStack QueryZustand + TanStack Query
Vercel

Backend

FastAPI (Python, async)FastAPI (Python, async)
PostgreSQL + SQLAlchemy ORMPostgreSQL + SQLAlchemy ORM
Role-based access control
Docker & Docker ComposeDocker & Docker Compose

Machine Learning

scikit-learn, LightGBM, pandas, numpy
Logistic Regression (Annual Model)
LightGBM + Logistic Regression Ensemble (Quarterly)

Infrastructure & DevOps

AWS ECS Fargate (API + workers)AWS ECS Fargate (API + workers)
AWS RDS (PostgreSQL)AWS RDS (PostgreSQL)
AWS ElastiCache (Redis)AWS ElastiCache (Redis)
AWS ALBAWS ALB
GitHub Actions → ECR → ECSGitHub Actions → ECR → ECS

My Role — End-to-End Ownership

Built entirely as a solo engineer, with full ownership across every layer:

  • •Designed the full system architecture and multi-tenant data model
  • •Built the complete FastAPI backend — auth, RBAC, tenants, orgs, users, jobs, predictions, and analytics
  • •Implemented async batch processing pipelines for CSV/Excel uploads
  • •Integrated ML inference workflows into production APIs
  • •Designed tenant-scoped data isolation across APIs, workers, and analytics
  • •Built frontend dashboards, forms, analytics views, and admin panels
  • •Implemented role-based access control across frontend and backend
  • •Set up Dockerized local dev environment and AWS production deployment
  • •Configured the full CI/CD pipeline (GitHub Actions → AWS ECS)

Outcome

AccuNode shipped as a production-ready multi-tenant ML platform with real-time and batch default risk predictions, scalable async job processing for large portfolios, secure role-based access and strict tenant isolation, and actionable analytics dashboards for financial decision-making.