I'm a 3rd semester AI/ML student building a closed-loop Brain-Computer Interface system that detects REM sleep and triggers haptic stimulation for lucid dream induction.
I call the long-term vision LUCID: Reality? — a dream interface where multiple people can share a collaboratively generated world during sleep. I call this technology Artificial Reality.
"When you are inside Artificial Reality, the question 'is this real?' becomes genuinely unanswerable."
My primary research project. A complete BCI pipeline: EEG signal → CNN-LSTM classifier → REM detection → safety interlock → haptic trigger Results so far (Sleep-EDF Cassette dataset):
| Experiment | Architecture | Accuracy | Kappa |
|---|---|---|---|
| 001 | SleepLSTM | 76.85% | — |
| 002 | CNN Spectrograms | 71.67% | — |
| 003 | CNN-LSTM Hybrid | 80.14% | 0.71 |
| 004 | CNN+Transformer | 79.00% | — |
| 005 | LOSO Validation | 77.17% | 0.67 |
| 006 | Full 153 subjects | TBD | TBD |
Tech stack: PyTorch · MNE-Python · scikit-learn · NumPy · SciPy
Machine Learning
Deep Learning architectures I've implemented from scratch:
- ✅ Neural networks (forward pass + backprop)
- ✅ CNNs — MNIST 98.47%, CIFAR-10
- ✅ LSTMs — sleep staging 80.14%
- ✅ Attention mechanisms
- ✅ Transformer encoder
- ✅ CNN-LSTM hybrid (BCI application)
Signal Processing
- EEG preprocessing (bandpass filter, epoch extraction)
- Sleep-EDF dataset (AASM standard staging)
- Real-time signal pipeline design
Tools
Closed-loop BCI for automated lucid dream induction. CNN-LSTM sleep stage classifier · 80.14% accuracy · Working haptic trigger. → github.com/anu-rag-007/PROJECT-07
Complete ML learning journey — from Python basics to transformers. 40+ LeetCode solutions, weekly notebooks, Kaggle projects (Titanic 78%), neural network from scratch. → github.com/anu-rag-007/Journey-to-the-BEST
Automatically detects soil dryness and waters the plant — no human effort needed. It is a great way to nurture plants keeping our resources in control and saving a hell lot of water. Arduino . Environment . ECE → github.com/anu-rag-007/smart_irrigation_system
Brain-Computer Interfaces → the hardware of LUCID EEG signal processing → reading the brain Deep learning for biosignals → understanding the brain AR / VR / XR → the delivery layer Neural rendering (NeRF) → generating dream worlds Generative AI → creating the content
Solving the Blind 75 list systematically. Arrays & Hashmaps ████████████ ✅ Stack & Queue ████████████ ✅ Trees & Recursion ████████████ ✅ Sliding Window ████████████ ✅ Linked Lists ████████████ ✅ Graphs ████████░░░░ in progress Dynamic Programming ████████░░░░ in progress
40+ problems solved · 7 core patterns
- 🔬 Experiment 006 — Expanding Sleep-EDF to 153 subjects
- 📄 Paper draft — IEEE TNSRE submission preparation
- 💡 Real-time pipeline — Muse S integration plan
- 📚 Week 9 of AI/ML self-study roadmap
LUCID: Reality? is a 10-15 year research initiative to create a shared dream interface — a technology I call Artificial Reality.
Phase 1 (now): BCI classifier + lucid induction Phase 2 (Year 2): Neural decoding + image generation Phase 3 (Year 3): 3D world generation + navigation Phase 4 (Year 5): Multi-user shared dream world Phase 5 (Year 10): Full Artificial Reality platform
The dream: when two people can enter the same generated world through sleep, indistinguishable from reality — that is Artificial Reality.
- 📧 Email — sharma.anurag0706@gmail.com
- 🔗 LinkedIn — linkedin.com/in/anurag-sharma-2a8b3b371
- 📊 Kaggle — kaggle.com/anuragsharma07112006
- 🧪 Project 07 — github.com/anu-rag-007/PROJECT-07
Building the foundation of Artificial Reality. One experiment at a time.