Abdullah
Khan
AI/ML Engineer _
1+ year shipping agentic AI, RAG, and voice systems in production — currently building RAG Systems and Voice agents at CR Equity AI.
Shipping AI systems
that handle real traffic.
Abdullah Khan
AI/ML Engineer
I'm Abdullah Khan, an AI/ML Engineer with 1+ year of experience specializing in agentic AI, RAG architectures, and production LLM & voice systems, with 6+ shipped projects handling real user traffic. I hold a BSc in Computer Science from The University of Faisalabad, CGPA 3.59/4.0.
At CR Equity AI, I've built and deployed 10 of 25 sub-agents for a CRM AI Assistant platform as FastAPI microservices on AWS, and I'm currently building two Twilio/ElevenLabs conversational voice agents. Before that, as an Associate AI Engineer at Xelent Solutions, I designed production multi-agent systems with LangGraph and FastAPI, including a multi-agent document editor and an automated code-generation pipeline.
I'm certified by DeepLearning.AI (Stanford) and IBM in deep learning, agentic AI frameworks (LangGraph, CrewAI, AutoGen, BeeAI), and generative AI. I'm now looking for my next role as an Associate/Junior AI-ML Engineer where I can keep shipping production systems.
GitHub Activity & Current Focus
Experience
- Built and deployed 10 of 25 sub-agents for the platform's CRM AI Assistant from scratch — including Dynamic Pricing, Compliance & Monitoring, and Performance Analytics — as FastAPI microservices on AWS, backed by NeonDB, Redis, and Apache Kafka.
- Automated ingestion and extraction of appraisals, rent rolls, and DSCR packages with LLM-powered document workflows, tuning prompt strategies across OpenAI, Claude, and open-source models.
- Rebuilt a Zoho SalesIQ chatbot's conversation flow and RAG pipeline to fix context retention across turns; that same knowledge base now also powers two in-development voice agents.
- Currently building two conversational voice agents (Twilio Programmable Voice, ElevenLabs Conversational AI) — a technical-support agent and a broker-facing sales agent — with FastAPI orchestration.
- Designed and delivered production agentic AI systems using LangGraph and FastAPI — multi-agent architectures with supervisor routing, parallel specialist agents, and planner-based orchestration beyond simple chatbot workflows.
- Built a multi-agent document editor (LangGraph, FastAPI, WebSocket streaming) with specialized writer, editor, and diagram agents, and an automated code-generation pipeline that turns a one-line idea into a complete, GitHub-pushed Python project.
- Integrated LLM APIs across OpenAI, Anthropic, and Groq, translating client requirements into scalable, production-ready systems.
- Prototyped early agentic workflows in LangChain/LangGraph that became the foundation for the team's later multi-agent architecture.
- Developed and maintained iOS application features and UI components alongside the core development team.
- Implemented, tested, and debugged application functionality, gaining hands-on experience with mobile development workflows and version control.
Technical Skills
Projects
01
OneScan — AI-Powered Multi-Cloud Security Scanner
Featured
Details
OneScan — AI-Powered Multi-Cloud Security Scanner
FeaturedA LangGraph pipeline that turns raw Prowler CLI findings into plain-language risk summaries and remediation guidance via the Claude API — unifying AWS, Azure, and GCP scanning into one tool.
- Sole backend developer: LangGraph + Claude API pipeline translating technical scan output into plain-language risk summaries
- Unified AWS, Azure, and GCP scanning with platform-specific credential validation (Boto3, Azure Identity SDK, Google Cloud SDK)
- Eliminated plaintext credential storage entirely — Fernet-encrypted, decrypted only in-memory during active scans
- Server-Sent Events pipeline streaming live progress for 2–20 minute compliance scans (GDPR, HIPAA, SOC2, CIS)
02
UniMind — RAG University Study Assistant
Details
UniMind — RAG University Study Assistant
Ask questions about your university notes using AI. Additive vector indexing, multi-format OCR, and resilient LLM routing with automatic fallback.
- Additive ChromaDB pipeline with MMR retrieval that preserves existing chunks, cutting search latency 40%
- Exponential-backoff retries with automatic fallback from GPT-OSS-120B to Qwen2.5-72B-Instruct, reaching 99.9% uptime
- Auto-routing OCR pipeline (PyMuPDF + Tesseract) extracting text from 100% of scanned and digital PDFs
03
Fraud Detection MLOps System
Details
Fraud Detection MLOps System
A real-time, event-driven fraud prevention MLOps system processing streaming transactions with Apache Kafka, FastAPI, and Gradio.
- Re-tuned the classification threshold to 0.265 via precision-recall analysis on 284,807 transactions (0.17% fraud rate), cutting false negatives 23%
- Integrated MLflow to log metrics, hyperparameters, and artifacts across every training run
- Containerized the API, Kafka broker, and Zookeeper with Docker Compose, cutting deployment setup to minutes
- Shipped FastAPI endpoints alongside a Gradio UI, live on HuggingFace Spaces with real-time Kafka streaming inference
04
Diamond Price Prediction
Details
Diamond Price Prediction
A production-grade MLOps system for predicting diamond prices with a modular pipeline tracked via MLflow.
- End-to-end MLOps pipeline: ingestion → transformation → training → deployment
- MLflow experiment tracking comparing multiple regression models
- FastAPI REST endpoint with Pydantic v2 input validation
- Gradio frontend with structured 4C diamond grading reference guide
05
CineMatch
Details
CineMatch
Content-based recommendation engine using cosine similarity across 5,000+ movies with real-time posters.
- CountVectorizer feature extraction on genres, cast, keywords & overviews
- TMDB REST API integration for live posters & release years using movie IDs
- Cached model inference with st.cache_resource for fast cold starts
06
HealthGuard AI
Details
HealthGuard AI
Three independent ML models predicting Diabetes, Heart Disease, and Parkinson's Disease.
- 3 separate SVM/Logistic Regression models with dedicated StandardScalers
- 1,266 patient records spanning 8–22 clinical features per disease
- Color-coded results with responsible AI disclaimers throughout
Certifications
Get In Touch
I'm looking for my next role as an Associate or Junior AI-ML Engineer. If you're building something with agentic AI, RAG, or production ML and need someone who can ship — let's talk.
Open to Opportunities
Currently an AI/ML Engineer at CR Equity AI, looking for my next role as an Associate or Junior AI-ML Engineer. Based in Pakistan, open to remote work globally.