Enterprise AI Engineer | GenAI Solutions Architect | RAG, VLMs, Multimodal Agents, Evaluation
I build retrieval-grounded AI systems for document intelligence, multilingual search, biomedical workflows, and secure enterprise deployments.
Recent work combines multimodal RAG, vision-language models, hybrid retrieval, LangGraph-style orchestration, benchmark-driven evaluation, AWS reference architectures, and governance controls such as citations, audit logs, and human review.
I am open to enterprise AI roles where the work needs engineering discipline, measurable retrieval quality, and deployment-aware architecture.
Best repositories to review first: CaseLens-VLM, SOAS RAG Evaluation, and Breast Cancer Multimodal AI.
Target roles: Applied AI Engineer, Enterprise GenAI Engineer, RAG/LLM Engineer, AI Solutions Architect.
| Work | Enterprise signal | Evidence |
|---|---|---|
| CaseLens-VLM | Multimodal document RAG over scanned pages with Qwen3-VL, hybrid retrieval, citations, audit controls, and AWS architecture mapping | Recall@5 improved from 0.035 metadata-only to 0.708 with Qwen3-VL + BM25/MiniLM hybrid retrieval on 339 DocVQA questions |
| SOAS RAG Evaluation | Bilingual RAG benchmark for culturally grounded English/Uzbek retrieval | Uzbek retrieval recall improved from 39% to 98% through corpus supplementation; Cohen's d = 2.91 |
| Breast Cancer Multimodal AI | Biomedical multimodal foundation-model benchmarking across pathology, genomics, clinical features, and mammography | CONCH V+C+G cross-attention C-index 0.609; Stage 1 AUROC 0.741; log-rank p = 0.005 |
| DialogXR / sovereign deployment work | Air-gapped enterprise AI deployment and multi-agent orchestration for secure environments | LangGraph orchestration, Llama 3.1 8B via Ollama, Lenovo ThinkSystem SR630 V4 deployment pattern |
| NVIDIA DLI instruction | Enterprise AI enablement and applied training delivery | Delivered RAG and multimodal AI agent workshops for academic and engineering audiences |
| Repository | What it demonstrates | Stack |
|---|---|---|
| caselens-vlm | VLM-assisted document intelligence, DocVQA retrieval evaluation, Slurm workflow, AWS reference architecture | Python, Qwen3-VL, BM25, MiniLM, Streamlit, Slurm |
| soas-rag-evaluation | Multilingual RAG evaluation, corpus engineering, statistical comparison of retrieval interventions | Python, retrieval eval, Hugging Face, Isambard |
| Breast-Cancer-Multimodal-AI | Biomedical AI benchmarking with multimodal survival prediction and governance-aware deployment design | Python, PyTorch, pathology foundation models, survival analysis |
| open-course-rag-benchmark | Open educational RAG benchmark with licensing-aware data handling and reproducible evaluation | Python, BM25, dense retrieval, OpenStax, pytest |
| cash-for-crash | Insurance fraud detection architecture using graph ML, RAG, and multi-agent workflows | Python, PyTorch Geometric, LangChain |
- Languages: Python, TypeScript, SQL
- RAG and Search: BM25, hybrid retrieval, vector search, FAISS, ChromaDB, Weaviate, OpenSearch patterns
- LLM/VLM Systems: Hugging Face Transformers, Qwen-VL, Llama, Ollama, OpenAI, Bedrock architecture patterns
- Agents: LangGraph, LangChain, tool routing, multi-agent workflow design
- ML: PyTorch, scikit-learn, survival analysis, pathology foundation models, graph neural networks
- Deployment: Docker, FastAPI, Streamlit, Slurm, Apptainer, AWS reference architectures, on-prem/air-gapped design
- Evaluation: Recall@k, MRR, AUROC, C-index, bootstrap confidence intervals, audit and failure taxonomy
- Production-oriented RAG and document intelligence design, not just demo chatbots
- Retrieval evaluation with explicit baselines, metrics, and failure analysis
- Experience with multilingual, multimodal, healthcare, education, and public-sector AI use cases
- Practical deployment thinking across AWS, HPC, on-prem, and air-gapped environments
- Governance-aware design: citations, human review, audit logs, limitations, and measurable quality checks
- Senior Applied AI Engineer roles
- Enterprise GenAI / RAG Engineer roles
- AI Solutions Architect roles
- Multimodal document intelligence and healthcare AI roles
- Remote or hybrid opportunities across the UK, EU, UAE, and enterprise AI teams globally
Building evaluated AI systems for real enterprise constraints: retrieval quality, provenance, governance, and deployment.