Arbaz Shaikh
Independent AI Systems Lead and ML/AI Engineer specializing in agentic AI, LLM-powered document understanding, and intelligent automation for procurement and compliance.
About
Independent AI Systems Lead and ML/AI Engineer specializing in agentic AI, LLM-powered document understanding, and intelligent automation for procurement and compliance. I build production-grade pipelines—from unstructured documents to structured outputs—while ensuring auditability and explainability.
I’ve led MVP delivery for tender-driven CapEx workflows, advanced transaction monitoring and sanctions intelligence, and shipped LLM assistants with privacy-preserving private inference stacks. I enjoy turning complex requirements into modular, API-ready systems that downstream teams can integrate into ERPs and operational platforms.
- 6 years' experience
- Joined July 2026
Industry experience
- Financial Services
- Healthcare
- Professional Services
- Computers & Electronics
- Software & Internet
Work experience· 6 roles
Delivered an MVP for tender-driven industrial/CapEx projects, improving execution speed by 2–10x across procurement stages (from document ingestion to structured output generation) and serving 20+ paid clients. Designed an end-to-end agentic backend with a modular rule engine to map unstructured document instructions to structured actions including BoQ/SOR, RFQ, vendor evaluation, technical evaluation, and tender review. Built document-understanding pipelines using PaddleOCR + LLM for parsing PDFs up to 1 GB and Excel files (1000+ rows) with conflict/missing-input detection across tender documents and BoQs. Exposed modular backend APIs for integration-ready consumption (e.g., ERP systems). Built a product hierarchy (7k+ products) and an ingestion layer to categorize products and standardize specifications.
First AI/ML hire at a seed startup; mentored 6 AI interns and 4 junior engineers; interviewed 10+ candidates. FactGuard (Transaction Monitoring): built an LSTM-based risk pipeline over 380M UK transactions, improving Recall@K by 3x versus baseline models; implemented a model explainability layer using SHAP plus LLM-based summaries to ensure SR 11-7 compliance. FactList (Sanctions Intelligence): built an LLM-driven feature that monitors EU press releases and extracts structured sanctions data with 96% accuracy; reduced data delays from days to minutes and transitioned from POCs to production with global clients. Built additional compliance systems including an inbox assistant for clustering email threads, extracting structured case data, and drafting responses; deployed a self-hosted private LLM inference stack using Qwen-2.5-14B with vLLM for data security. Also delivered KYC document extraction (OCR+LLM), address verification, merchant risk monitoring, adverse media knowledge graph pi
Built a RAG-based medical AI chatbot and explored LLM usage for clinical coding in a pre-seed startup environment.
Education
Hire Arbaz Shaikh today
Post a job and invite Arbaz, or contact them directly. No hidden commission.



