ML Lead. I build agentic AI systems for healthcare — the kind that query 1,200-table clinical data warehouses, extract structure from messy clinical notes, and actually run in production.
I believe in shipping over theorizing, agents over dashboards, and building tools I'd want to use myself.
Blog • LinkedIn • singhrohan@outlook.com
- AgentHub — GitHub is for humans. AgentHub is for agents. Infrastructure for autonomous research agents. (Go)
- clin-summ — Clinical text summarization by adapting LLMs for healthcare documentation. (Python)
- clinExtract — Structured extraction from unstructured clinical narratives. (Python)
- synthmedreports — Synthetic medical data generation — for when you can't use real patient data. (Python)
- freeWispr — Free, local, privacy-first dictation for macOS. (Swift)
- gpt-obsidian — Turns 1,373+ ChatGPT conversations into a searchable Obsidian vault. (Python)
- 12 Patterns I Learned Building an Agentic AI That Queries a 1,200-Table Data Warehouse
- From Tokens to Tensors: Inference Engineering
- Attention Is All You Need
- Importing 1,373 ChatGPT Conversations Into Obsidian
More at rohansweblog.com