Shahiryar S.

I design and ship production AI systems for regulated organizations, such as banks, financial groups and government ministries, where data can't leave the building and every answer has to be traceable. I'm a Google Cloud certified architect, a Microsoft Certified Azure AI Engineer, and a Google award-winning consultant in the public and private sectors (2025-26). I work as a Senior Consultant at a digital transformation consultancy, where I'm the technical lead for AI applications and the data team on client work in finance and manufacturing. I've built ML systems professionally since 2021, starting in Japan, and I take AI from proof of concept to production and stay through deployment. RECENT WORK • Fully offline LLM platform for a financial aggregator. Qwen2.5-32B (4-bit AWQ) runs on NVIDIA L4 and RTX Pro GPUs, served with vLLM, with LoRA/QLoRA fine-tuning on domain data and a multi-agent system on top. No data leaves the client's network. • Cloud transformation for a financial conglomerate, as lead architect for one year. I ran discovery across 19 business units, designed the multi-cloud target architecture, and led the live migration of workloads from Azure to Google Cloud. • Government-wide AI and data infrastructure (ongoing). I'm building the core platform for AI/ML workloads across ministries, with strict data governance. • Secure AI gateway and ISO/GDPR compliance design for a German software company, rated 5.0. • Semantic search in Japan: BERT embeddings over 400,000+ records with FAISS-based knowledge graphs. I also led the ML deployment team. WHAT I CAN BUILD FOR YOU • Agentic AI and multi-agent workflows with tool calling, human approval steps and audit trails, on Vertex AI / Gemini Enterprise or Azure OpenAI • RAG and enterprise search that answer from your documents with citations, plus retrieval evaluation so you know the answers are right • On-prem or private-cloud LLM deployment, covering model choice, GPU sizing, quantization, vLLM serving and fine-tuning • Cloud backends for AI on Google Cloud (Cloud Run, BigQuery, Apigee) and Azure • AI architecture reviews and adoption roadmaps for teams that want an independent second opinion Stack: Python, FastAPI, Vertex AI, Gemini, Azure OpenAI, LangChain, LlamaIndex, vLLM, BigQuery, Cloud Run, Apigee, pgvector, Docker, Terraform How I work: we start with a short discovery call. Then I write a plan with fixed milestones and ship working software early, so you can give feedback as we go. I also have a background in corporate strategy and international legal studies, which helps when the work involves compliance, data governance or risk. I'm available 14+ hrs/week, with overlap for UK and US working hours.