Lawrance Reddy · Microsoft MVP, AI Services · Durban
Building AI that ends up in the field.
I'm Lawrance Reddy. Microsoft MVP in the AI Services (Foundry) category, CTO at Vaxowave, based in Durban, South Africa.
Most of my work sits where Microsoft AI Foundry meets a real-world problem that would otherwise be expensive or impossible to solve. Two of those problems — verifying clean-energy and biodiversity outcomes for funders in rural KwaZulu-Natal, and giving South Africans access to government services in the language they actually speak — are the reason this blog exists.
The articles are practical and opinionated. Azure is specifically how I build, but the patterns generalise: grounding, tamper-evident ledgers, multilingual agents, offline-first.
What I build now
- ConservAxion — an Azure AI Foundry-backed platform that verifies clean-energy generation and biodiversity outcomes for funders and conservation partners. The idea and the conservation domain are Craig Beech's, the founder of ConservAxion. I'm the Azure architect building the technical platform behind his vision. Pilot stage in KwaZulu-Natal.
- Pfula — a conversational assistant that walks South Africans through SASSA, Home Affairs, SARS, UIF, municipal, deeds, and CIPC processes in isiZulu and English. Running on Azure AI Foundry + Foundry Agent Service.
Earlier work
In 2019 I was part of the team behind the Microsoft × Peace Parks Foundation rhino anti-poaching project at the Hluhluwe-iMfolozi reserve, run in partnership with Ezemvelo KZN Wildlife. Motion-activated camera traps fed an Azure Machine Learning pipeline that filtered the daily photo deluge for humans and vehicles, and routed real-time alerts to the ranger reaction units covering the 96,000-hectare park. Stack at the time was TensorFlow + Keras (SSD, Faster R-CNN, YOLOv3) on Azure ML Compute with GPU, behind Azure Functions and Postgres, served from Azure's South Africa region.
ConservAxion is the next chapter of that work — same region, same intent, different stack: Foundry agents and tamper-evident attestations instead of computer-vision models. The case study from the original project is on YouTube (Microsoft Developer · 2019).
Community
I co-organise Data & AI Community Day Durban (four free, in-person events a year) with Niels Berglund. Recent editions have run at Coastlands Umhlanga and Richfield College Umhlanga. The standing programme: 50-minute deep dives, 20-minute lightning talks, a panel discussion, and a structured first-time speaker track.
How to reach me
LinkedIn is the cleanest channel. For security or abuse reports, see security.txt.