Co-Founder & CTO @ Twinning Labs · Ex-Microsoft (Copilot) · AI Researcher
99% of the market optimizes after the fact. We took the other path: simulate the outcome before you spend a dollar.
At Twinning Labs, I'm building AI digital twins of entire enterprise populations so Fortune 50 companies can test decisions instantly, instead of running 2-3 week A/B cycles. Before that, I shipped production AI at Microsoft on the Copilot team and built ML/DL systems with IISc and Amazon.
- Enterprise Simulation Engine: Living world models that replicate business populations and simulate decisions before they ship
- Digital Twin Research: Identity and memory modeling, synthetic population generation, and decision governance
- ML Infrastructure: Experiment frameworks, reproducibility tooling, and CI/CD for ML pipelines
| Where | What |
|---|---|
| Twinning Labs | AI digital twins, enterprise simulation, decision governance |
| Microsoft | AI software engineering on Microsoft Copilot |
| IISc CST Department | ML/DL systems for plasma arc detection and analysis |
| Amazon | Intent classification and slot filling for conversational AI |
ML/AI: PyTorch · TensorFlow · Hugging Face · scikit-learn · LangChain · ONNX
Infrastructure: Docker · Kubernetes · AWS · Azure · FastAPI · PostgreSQL
Languages: Python · TypeScript · C++ · Go
These repos showcase techniques from my research and industry work, not proprietary code.
| Project | Description |
|---|---|
| plasma-arc-detection | ML pipeline for plasma arc detection, based on IISc research |
| intent-slot-filling | Joint intent classification and slot filling NLU system |
| ml-experiment-template | Opinionated ML experiment framework with config management and reproducibility |
| paper-implementations | Clean PyTorch implementations of key papers in NLU, medical imaging, and CV |
Most AI companies optimize after the fact. I think the real leverage is in simulation: knowing what will work before you launch. That's the thesis behind Twinning Labs, and it's the thread running through everything I build.
If you're working on simulation, digital twins, or enterprise AI, let's connect.
