Abdur Rafay

Software Engineer - AI/Backend

Kachyng-RokketAI • Tiempo completo

Feb 2024 - Feb 2026 • 2 yrs

AI & LLM Infrastructure Engineered an "Agentic Query" layer within a Data Fabric, utilizing LLM orchestration to intelligently interface with multiple database schemas and generate precise SQL responses. Built a Guardrail and PII Detection service using the Presidio library and custom classifiers, ensuring enterprise-grade data privacy for LLM inputs and outputs. Optimized LLM deployment by managing quantized large language models within Kubernetes clusters. Automated prompt engineering workflows by integrating OpenAI APIs with a Milvus vector database for high-speed similarity searches and background task-driven prompt insertion. Backend Architecture & Distributed Systems Architected a multi-tenant Revenue and Commission Engine that automates complex financial metrics, including brand holdbacks, tax percentages, and dynamic dealer fees. Scaled a Serverless Monorepo using NPM Workspaces to consolidate 15+ microservices, reducing deployment package sizes and ensuring version consistency across the ecosystem. Optimized mission-critical API performance, achieving a 60% reduction in latency (from 90s to 36s) through targeted database indexing and TypeORM query optimizations. Modernized legacy data layers by migrating from direct database access to a centralized, API-driven model using jsonb for faster complex attribute retrieval. Designed a high-concurrency "Orders Service" using AWS Lambda and ElastiCache, enabling atomic order generation and reducing database contention during high-traffic events. Cloud Infrastructure, Security & DevOps Ensured PCI-DSS compliance by engineering a secure multi-tenant session management system utilizing JWT (HS256/RS256) and AWS Secrets Manager for payment token hashing. Centralized Identity Management by migrating authentication logic for the entire service ecosystem to AWS Cognito, supporting multi-tenant resource management. Standardized "Boot-Up" orchestration via a unified Docker Compose framework that manages inter-service depende

Software Engineer - Backend

Auxin Security • Tiempo completo

May 2022 - Jan 2024 • 1 yr 8 mos

Backend & Distributed Systems Designed a scalable, multi-channel notification engine utilizing WebSockets (Socket.IO) and asynchronous processing (Celery/Redis) to deliver real-time alerts with 100% delivery reliability. Built a distributed background task architecture using Celery Beat to automate high-concurrency cloud inventory syncs and organization-level data uploads to Milvus. Streamlined service discovery and reduced latency by configuring an Nginx reverse proxy for 10+ microservices, implementing path-based routing and regex patterns for seamless API documentation. Refined data persistence layers by migrating cloud inventory tasks to a factory pattern, optimizing database connection flows for high-concurrency environments. AI & LLM Engineering Designed a secure AI-driven dashboard API using Large Language Models (LLMs) for vulnerability prompt validation and automated security insight generation. Optimized Retrieval-Augmented Generation (RAG) performance by integrating Vector Databases (Milvus/Chroma) and fine-tuning text embedding parameters, significantly increasing the accuracy of AI findings. Developed advanced conversational AI agents using LangChain, incorporating complex memory management, multi-step reasoning chains, and context-aware chat capabilities. Engineered a custom token-tracking and exchange API for OpenAI and Alpha-AI, enabling granular observability and cost-efficiency management of LLM usage across the platform. Security & DevSecOps Engineered a comprehensive vulnerability management backend to manage multi-platform cloud inventory (AWS, GCP, Azure) and integrated it with DefectDojo and Jenkins for automated finding management. Developed a high-performance Python-based CLI utility that enables developers to perform SAST and Infrastructure-as-Code (IaC) scans on local directories and GKE helm charts. Expanded platform security coverage by implementing specialized parsers for Kubernetes, Secret Scanning, and AI-powered Static Analysis. Fron