I'm an AI Engineer based in Islamabad, Pakistan, with 3+ years of experience designing, building, and deploying production-grade LLM systems — from agentic multi-agent orchestration pipelines to retrieval-augmented generation (RAG) architectures and scalable AWS infrastructure for enterprise clients.
- 🔭 Most recently led architecture & engineering of GovHub — a production AI platform ingesting thousands of SAM.gov federal contracting opportunities daily through a LangGraph multi-agent pipeline with automated compliance checking, scoring, and executive summary generation.
- 🧠 Deep expertise across the full AI stack: multi-agent system design with LangGraph StateGraphs, tool/function calling, hybrid vector search, LLM observability with Langfuse, prompt engineering, and end-to-end AWS deployment.
- 👯 Looking to collaborate on AI agents, RAG systems, and intelligent automation platforms.
- 💬 Ask me about LangGraph, RAG pipelines, FastAPI, vector search, and LLM deployment on AWS.
- 📫 Reach me at malikmuhammadkashif414@gmail.com
GovHub — AI Government Contracting Intelligence Platform (Lead AI Engineer, Marsons Media)
Production SaaS platform ingesting thousands of SAM.gov opportunities daily via a 6-agent LangGraph pipeline (
llm_match,document_parser,compliance_checker,score_calculator,opportunity_orchestrator,company_match), delivering ranked opportunities with compliance reports and win-probability scores. Deployed on ECS Fargate, Lambda, RDS, and S3 with full GitHub Actions CI/CD.
Intella Part AI — Industrial RAG System (Codeaza Technologies)
Production RAG system adopted across 30+ industrial brands, indexing 5,000+ technical manuals with Voyage AI embeddings and Elasticsearch, delivering sub-second p95 query latency.
Real Estate Property QA Chatbot (Codeaza Technologies)
Hybrid retrieval architecture (Qdrant + BM25 + reranking) achieving ~50% improvement in search relevance, served via a FastAPI backend with <200ms p95 response time.
🟢 GovHub — AI Government Contracting Intelligence Platform
- Led full architecture & engineering of a production AI SaaS platform ingesting thousands of SAM.gov opportunities daily, running a multi-stage LangGraph pipeline with compliance reports and win-probability scores.
- Designed a 6-agent LangGraph StateGraph orchestration system (
llm_match,document_parser,compliance_checker,score_calculator,opportunity_orchestrator,company_match) via an agent registry pattern with runtime dependency injection. - Built a two-stage AI pipeline: real-time GPT-4.1-mini quick-match scoring, plus on-demand full RFP/SOW parsing and BLUF summary generation via Claude's 200K context window.
- Engineered the SAM.gov ingestion pipeline (paginated fetch → PostgreSQL upsert → async S3 streaming → Lambda → ECS Fargate dispatch) and a multi-format document intelligence layer (PDF, Word, Excel, HTML).
- Integrated Langfuse for full LLM observability — traces, token cost, and latency across all agents.
- Deployed on AWS (ECS Fargate, Lambda, RDS, S3, ECR, Secrets Manager, private-subnet VPC) with a GitHub Actions CI/CD pipeline for zero-downtime rolling updates.
🔵 Intella Part AI — Industrial RAG System (30+ Brands, 4,000+ Manuals)
- Architected and shipped a production RAG system adopted across 30+ industrial brands for natural-language Q&A over technical documentation.
- Built an ingestion pipeline for 5,000+ manuals using Voyage AI embeddings and Elasticsearch, hitting sub-second p95 query latency.
🟣 Real Estate Property QA Chatbot
- Designed a hybrid retrieval architecture (Qdrant + BM25 + reranking), improving search relevance by ~50% over baseline keyword search.
- Built a FastAPI backend with <200ms p95 response time; owned the project end-to-end from discovery to AWS deployment.
🟠 Sentimantle — ML Sentiment Analysis Platform
- Maintained production ML inference pipelines on AWS ECS, sustaining 99%+ uptime.
- Cut API response times by 40% through MySQL query rewriting and index tuning.
🔴 Document QA Chatbot & LLM Benchmarking Tools
- Built a LlamaIndex-based RAG chatbot for semantic Q&A over uploaded documents.
- Developed LLM Wars, a GPT vs. Gemini output-quality benchmarking framework.
- Built Echo App, a Node.js backend integrating an AI image model for automated social content generation.
🎓 B.E. Computer Software Engineering — National University of Sciences and Technology (NUST), Islamabad
AI / Agents
RAG & Vector Search
Backend
Frontend
Cloud & DevOps
Databases
ML / Data
