I'm a civil engineer who builds AI tools for ageing infrastructure: turning drone footage, LiDAR scans and engineering specs into something an engineer can actually act on.
Co-founder of CorrSens
Structural audits, NDT testing and corrosion protection for critical infrastructure, developed at IIT Madras.
- Bridgescan turns bridge point clouds into reviewable BIM elements
- InfraTWIN is a digital-twin platform for inspection data
| Where | What | |
|---|---|---|
| now | Indian School of Business, Hyderabad | IVI program |
| before | IIT Madras | Computer vision research |
| start | NIT Calicut | B.Tech |
| Project | What it is |
|---|---|
| Bigram Language Model | A character-level language model in PyTorch, built two ways: by counting letter pairs and as a one-layer neural net. Trained on Indian and English names. ▶ Try the Name Machine |
| nrjNET | My first neural network, built from scratch in pure Python. Autograd, backprop and an MLP, with gradients verified against PyTorch. |
| Civion | A compliance engine for IS 456:2000 concrete specs. Upload a spec PDF, get an audit, chat about it and generate a corrected version. |
| Gaussian Splatting Demo | Gaussian splatting, a new kind of 3D modelling, applied to a bridge in Kerala |
| Construction Object Detection | YOLOv8 material detection for construction progress monitoring |
Also: YOLOv8 · point clouds (LAS) · BIM / IFC · photogrammetry · Gaussian splatting
Deep learning from first principles: building every piece by hand before using the frameworks. Started with nrjNET, now working through language models with a bigram name generator.
turning visual data into engineering intelligence