Created a B2C unstructured document processing SaaS. Provisioning NVIDIA-runtime capable bare metal K8s clusters to reduce our operational costs, allowing for ultra-competitive service pricing

Developed an in-house usage-based metering system handling 100,000+ usage events per second for our Saturn document extraction service.

Provisioning and optimising GPU infrastructure (in cloud and on-prem) for serving open source Transformer and OCR models, achieving ~30% cost reduction compared to Huggingface

Others: vectorised BM25 retrieval algorithm with sub 100ms response time under load, developing Semantic Router OSS library, setting up payments & auth for Aurelio Saturn

Can I run and serve an AI Newspaper entirely from my living room, using just 3 dumpster-dived HP Prodesks and clever caching? Yes, Trending on Weibo performs automatic trending topic discovery, topic clustering, agentic scraping and agentic deep-research, orchestrated using Temporal.

Serving a 15.7k monthly page views website with a 99 Lighthouse Score from my residential network, using Cloudflare tunneling & OpenNext.js

Overhauled company Infrastructure into a global, zero VPC peering, zero trust, self-service, private networked Engineering Platform. Improved dev team iteration speed while maintaining Platform team approval over core infrastructure changes

Adopted GitOps-based CI/CD strategy, with ArgoCD at its core, providing dev teams with DevSecOps-compliant base templates (for PCIe & ISO-27001)

Introduced Istio Service Mesh across our globally distributed private K8s clusters, for cross-service encryption using mTLS, out-of-the-box telemetry & service discovery. Zero-downtime migration onto the mesh

Helmed creation of Incident Response guides within company, and adoption of deeper instrumentation using Datadog

Reduced company Elasticsearch spend by 64%, using cold-node advanced index lifecycle management

Created an idempotent Python data ingestion pipeline for FMCG retailer data, tracking data quality and change in feature importance. Improved batch job throughput by 30x. Solved supply chain issues by creating a product allocation algorithm, achieved a 40% uplift for shelf fill rates

Hired and managed a team with two junior Data Engineers; wrote Engineering onboarding guide