Clinical Decision Support ToolFull-stack web app helping healthcare professionals manage patient records and analyze treatment options with AI assistance. End-to-end production AI system.
React
TypeScript
Express
OpenAI API
Hi, I'm John Legaspi
Senior Data & Platform Engineer in London. I design AI-native data platforms and agentic LLM workflows.About
The short version
I specialize in high-scale data infrastructure and agentic AI workflows. At McLaren Construction I've been pioneering LLM-orchestrated engineering — multi-agent frameworks that translate technical requirements into production Microsoft Fabric workspaces, and that reduced our Oracle-to-lakehouse migration cycle from 20 days to 5.My background spans platform engineering, MCP-based autonomous systems, Python/PySpark orchestration, and Medallion architectures. I focus on merging engineering rigor with AI governance to drive measurable business impact — and I'm looking for Staff or Lead roles to own AI-infrastructure strategy and next-generation data architecture.
75%Cut on Oracle→lakehouse migration cycle
90%Reduction in cloud compute spend
20+Production pipelines, millions of records/day
99.9%Airflow pipeline reliability
Experience
Where I’ve worked
Senior Data & Platform Engineer · McLaren ConstructionJul 2024 – Present · Promoted from Data Engineer in Dec 2025
- Architected an LLM-agent framework that cut the Oracle-to-lakehouse migration cycle from 20 days to 5 days — a 75% reduction with full data lineage preserved.
- Cut ingestion times from 1 hour to 2 minutes using asyncio concurrency and optimized PySpark; reduced cloud compute spend by 90% via compute/IO separation.
- Designed a multi-agent orchestration framework that turns technical requirements into production Microsoft Fabric workspaces and artifacts.
- Built an API ingestion framework with circuit breakers and exponential backoff, eliminating cascade failures and throttling errors across upstream integrations.
- Delivered 20+ production pipelines processing millions of records daily, with Pydantic schema validation and CDC-based incremental loading.
- Established Bronze/Silver/Gold Medallion structure in Delta Lake — reduced ad-hoc data requests by 40%.
- Built Azure DevOps CI/CD pipelines enforcing 100% code review coverage; shortened release cycles from weeks to days.
- Production-grade Airflow DAGs with custom failure callbacks and SLA enforcement — 99.9% pipeline reliability.
- Analytics and custom KPIs contributed to McLaren winning Digital Contractor of the Year 2026.
Technical Project Manager · Asante MediaJan 2024 – Jul 2024
- Bridged business stakeholders and engineering teams, translating client requirements into functional specs and technical documentation.
- Owned planning, execution, and monitoring of multiple concurrent web projects — delivered on time and on budget.
Front-End Developer · Asante MediaAug 2021 – Jan 2024
- Engineered responsive components and integrated Algolia search-as-a-service APIs for high-performance retrieval and real-time interaction tracking.
- Built reusable, extensible web components using AEM’s Java-based component framework in cross-functional Agile teams.
Education & certificationsBSc Environmental Science — Queen Mary University of London (2:1)Data Analyst in SQL — DataCamp
Projects
Selected work
WhatsApp Data Extraction PipelinePython CLI that parses WhatsApp chat exports and extracts structured records via the Claude API. Few-shot prompting, prompt caching, batched & resumable calls, YAML-configurable schemas, rapidfuzz-based deduplication.
Python
Claude API
Pydantic
YAML
rapidfuzz
Skills
Tools of the trade
Languages & frameworks
PythonPySparkasyncioPandasPydanticSQLT-SQLPL/SQLPostgreSQLTypeScriptReactNext.js
Data platforms
Delta LakedbtMedallionCDCStar SchemaSnowflake Schema
AI / LLM engineering
LLM agent orchestrationMulti-agent frameworksMCP serversFew-shot promptingPrompt cachingRAG systems
Cloud & DevOps
Azure FabricAzure SQLAzure DevOpsGitHub ActionsEntra ID
Orchestration
Apache AirflowCustom operatorsDAGsSLA enforcementFailure callbacks
BI & analytics
Power BIDAXPower QueryKPI designCanvas Apps
Contact
Let’s talk
Whether you're hiring for a Staff or Lead role, want to discuss AI-native data architecture, or just want to say hello — I'd love to hear from you.
Reach me directly at jmlegaspi1@outlook.com