From POC to scale: long-running AI workflows
How we represent human work as a graph, run it with LLMs, and what changes when the workflow has to survive hours, days, and thousands of concurrent runs.
I'm Sachin Kahawala, a Senior Data Scientist & AI Architect based in Melbourne, AU. I design and ship production AI systems: LLM pipelines, multi-agent workflows, and the evaluation harnesses that keep them honest. My background is a PhD in AI and 6+ years across research and industry.
Latest writing
All posts →Featured projects
All projects →Planning Compliance Automation Platform
Lead AI Architect · Apr 2025 - Present
Reduced manual planning review time by >70% with cited, audit-ready outputs.
- Azure OpenAI
- Azure AI Search
- FastAPI
- React
- LangGraph
- DeepEval
- +4
Enterprise Knowledge Retrieval Assistant (RAG)
AI Architect · Oct 2024 - Feb 2025
<1.8s latency RAG assistant; ~30% reduction in Tier-2 escalations.
- Azure OpenAI
- Azure AI Search
- Python
- LangChain
- Azure Functions
- Blob Storage
Real-Time Voice Agent with CRM Integration
AI Architect · Feb 2025 - Mar 2025
Live voice agent prototype with mid-conversation CRM read/write via MCP tools.
- Vertex AI
- Gemini real-time
- LiveKit
- MCP
- Python
- FastAPI
- +2
Get in touch
Open to interesting AI/ML problems and collaborations. Reach me on LinkedIn, GitHub, or S.Kahawala@latrobe.edu.au.