MU
MUHAMMAD USMANSOFTWARE ENGINEER • FLUTTER & AI
Phebsoft EcosystemBackend & AI Infrastructure Engineer2024
Complimate AI & Cani-DoIt
Enterprise AI Compliance & Legal Verification Assistant
Doc Processing100+ Pages/min
Accuracy98.7%
API Latency180ms
StackPython / RAG
THE ENGINEERING CHALLENGE
Legal and compliance teams spent tens of hours manually verifying cross-border clauses against dynamic regulatory databases.
THE ARCHITECTURAL SOLUTION
Created an asynchronous RAG processing pipeline using LangChain, OpenAI embeddings, and ChromaDB vector store hosted on Dockerized Linux servers with Nginx reverse proxy.
SYSTEM ARCHITECTURE & DATA FLOW
Enterprise Document Vectorization Pipeline
High-security text extraction, semantic chunking, and similarity ranking against legal codices.
01 // STAGEDocument Ingestion (PDF / DOCX)
02 // STAGEPyPDF / OCR Parsing & Clean Text Extraction
03 // STAGEOpenAI Embedding Generation
04 // STAGEChromaDB Vector Retrieval
05 // STAGELLM Synthesis with Strict Citation References
KEY ENGINEERING HIGHLIGHTS & SOLVED BOTTLENECKS
Strict Citation & Hallucination Guardrails
Designed prompt validation layers requiring exact paragraph page citations for every compliance assessment.
Python • OpenAI API • Vector Search
Dockerized Scalable Microservices
Orchestrated multi-container deployment using Docker Compose, PM2, and Nginx with automated SSL renewal.
Docker • Nginx • Linux Ubuntu • GitHub Actions
TECHNOLOGY STACK & INFRASTRUCTURE
PythonFastAPI / FlaskOpenAI APIChromaDBDockerNginxLinux