AI Solutions Architect & Agentic Transformation Lead
Deloitte
London, UK
Technical pre-sales lead and proposition builder for AI customer engagements, working through rapid prototyping and architecture.
Prototype-led pre-sales
Ran an outcome-based prototyping motion across client accounts, building working, interactive proof-of-concept applications in days with small engineering pods using Python (FastAPI), SvelteKit, ADK on Agent Runtime, LangGraph and Cloud Run, in place of multi-week slide cycles. Validated prototypes progressed into live customer pilots supporting 500+ active users, helped close sales cycles and secured follow-on paid pilot phases.
Text-to-SQL agents with semantic caching
Architected text-to-SQL agent workflows over enterprise BigQuery data estates and implemented semantic vector caching with Firestore and embeddings, removing repeated inference costs and giving low-latency document and database grounding.
Agent factory templates
Architected open-standard GitHub template repositories for an agent factory bootstrapper and a pre-sales vibe-coding demo harness, with zero-touch Terraform IAM, automated CI/CD, Pydantic tool-schema linting and custom MCP servers, so sales engineers and delivery teams could bootstrap grounded, tool-using agents in hours.
LLM-as-a-judge evaluation
Lead technical architect and advisor on client AI initiatives: implemented LLM-as-a-judge evaluation patterns to align outputs with client value definitions, wrote structured tool-calling schemas in Python, and configured authentication and runtime state management.
Multi-agent platform architecture and sizing
Designed the cloud architecture and runtime pricing model for an autonomous multi-agent procurement platform, modelling data throughput, storage and inference requirements, and secured $250k in vendor co-funding to move the client from prototype to production delivery.