Case StudyFinTech / SBA Lending

SBA
LOANS

HQ

Multi-tenant SBA lending SaaS for U.S. acquisition financing — AI document processing with AWS Textract OCR and Google Gemini, stage-based automation on AWS Lambda, and DocuSeal-powered e-signature flows. Higher revenue, lower ops drag, fully visible pipelines.

Role: Full-Stack Developer & Team LeadTimeline: 2024Stack: Next.js (Pages), Django REST, PostgreSQL / RDS
01 / Overview
"One system for the whole loan lifecycle — brokers, borrowers, and lenders in sync."

SBA Loans HQ is a U.S. consultancy focused on acquisition financing; the business lived in email threads, spreadsheets, and a patchwork of third-party tools with little shared visibility. I led delivery of a production-grade SaaS platform that centralizes deals, documents, and communication in a single workflow. The result is a scalable, multi-role product that replaces manual follow-ups with stage-based automation and gives every stakeholder a live view of progress.

My Role

I owned full-stack architecture and delivery as team lead: Next.js front end, Django REST API, PostgreSQL on RDS, and AWS for compute, storage, and document processing. Migrated document storage and automation off brittle connector stacks into first-party services; aligned eng with loan operations so business rules shipped as reliable product behavior. This work was delivered with Arithmiks as the consulting partner to SBA Loans HQ.

02 / System Architecture

hover to explore

FRONTEND01API LAYER02ASYNC LAYER03DATA STORE04AI PIPELINE05HTTPSHTTP/2PROXY_PASS :8000TASK.DELAY()PSYCOPG2 ORMPRE-SIGNED URLBRPOPSAVE RESULTSMULTIPART PUTBOTO3 CLIENTOCR CHUNKSCLIENTBrowserRedux · JWTCLIENTNext.js 16Pages Router · SSRSERVERNginxTLS · CertbotSERVERDjango + DRFJWT · RBAC · v0/SERVICERedisBroker + ResultsSERVICECeleryWorkers · BeatDBPostgreSQLpsycopg2 · MatViewsSERVICEAWS S3boto3 · 3 BucketsAPIAWS TextractOCR EngineEXTERNALGemini 1.5google-genai SDK
client
server
service
database
api
external
Async / external

SBA LOANS HQ system architecture: Browser (Redux · JWT); Next.js 16 (Pages Router · SSR); Nginx (TLS · Certbot); Django + DRF (JWT · RBAC · v0/); Redis (Broker + Results); Celery (Workers · Beat); PostgreSQL (psycopg2 · MatViews); AWS S3 (boto3 · 3 Buckets); AWS Textract (OCR Engine); Gemini 1.5 (google-genai SDK).

Connections: Browser → Next.js 16 (HTTPS); Next.js 16 → Nginx (HTTP/2); Nginx → Django + DRF (proxy_pass :8000); Django + DRF → Redis (Task.delay()); Django + DRF → PostgreSQL (psycopg2 ORM); Django + DRF → AWS S3 (pre-signed URL); Redis → Celery (BRPOP); Celery → PostgreSQL (save results); Celery → AWS S3 (multipart PUT); Celery → AWS Textract (boto3 client); AWS Textract → Gemini 1.5 (OCR chunks).

03 / Process
01

Problem Framing & Consolidation

Mapped how brokers actually worked: fragmented email, inconsistent file intake, no single pipeline view, and recurring manual nudges across the loan lifecycle. Defined a product boundary that replaced ad-hoc spreadsheets and generic CRMs with structured deal records, role-based access, and audit-friendly document handling — grounded in how SBA consulting actually closes deals.

02

AI & Event-Driven Automation

Layered in AI and OCR so the platform does more than CRUD: Gemini for summarization, intelligent file naming, and lead scoring; AWS Textract to pull structured data from complex financial PDFs. Built a custom automation path with stage-based bundles, outbound email via Postmark, and serverless hooks on Lambda so milestones drive the right requests and updates without Zapier-style glue.

03

Platform Delivery & Observability

Shipped a multi-tenant experience for partners, borrowers, lenders, and external requesters: Kanban and table views for the pipeline, an S3-backed vault that requests the right documents per stage, a memorandum section for underwriting narrative alongside live collaboration, DocuSeal for in-portal e-sign, and threaded per-deal chat to cut email noise. Added a super-admin analytics surface (ApexCharts) plus Mixpanel and GA4/GTM for product telemetry, and a database-backed error logger with full user context for production debugging.

04 / Results
+25%
Revenue Growth
↓ 60%
Operational Overhead
↑ 30%
Processing Speed
100%
Workflow Visibility
05 / Tech Stack
Next.js (Pages)Django RESTPostgreSQL / RDSGeminiAWS TextractAWS EC2 · S3 · LambdaPostmarkDocuSealMixpanelGA4 / GTMApexCharts
Hire me on Contra