True seeing is not with these two tiny windows of Maya…
It is VISION awakening, then cosmic PERCEPTION flooding the third eye like monsoon rain,
and finally… the grand INTERPRETATION — when the drunken monkey mind surrenders and whispers:
“Oho! Now I understand… everything was always Brahman wearing funny costume!”
I am Subhansh Malviya, the humble seer of pixels and gradients.
In the great maya of production
, I do not chase shiny new transformers.
No-no. I take the ancient, efficient mantras already existing… and make them dance faster, cheaper, stronger — sometimes I even whisper a new mantra myself, born purely from real-world suffering, not from leaderboard puja.
Quantization? Finetuning? These are but small toys for small minds.
I sit in deep samadhi at the feet of mathematics herself —
understanding the sacred geometry of architectures,
the karmic dance of loss functions,
the cosmic statistics that govern all things visible and invisible,
and the illusory benchmarks that fool the unenlightened.
Come, child… drop the ego of novelty.
Seek the eternal intuition behind the code.
Only then true vision awakens. 🕉️
Ahhh, listen carefully with open third eye…
My sacred research domain is Computer Vision – the divine art of teaching machines to see Maya of this world!
I chant mainly through PyTorch and the holy Transformers… sometimes LangChain also joins the puja.
For data and MLOps sadhana I worship Pandas and NumPy as holy cows, Matplotlib for drawing mandalas of loss curves, Docker as my sacred pot, FAISS for lightning-fast vector darshan, while TensorBoard and Weights & Biases are my jyotirlinga tracking every breath of the model.
Daily I speak the mantras of Python, offer prayers through SQL, walk the path of Git and Linux like true sannyasi, and meditate deeply inside Jupyter notebook samadhi.
When time comes for worldly manifestation, I deploy on AWS – the cloud Himalaya, or gracefully on Vercel, Render, and Hugging Face Spaces – like lotus flowers blooming without effort.
Aum Shanti Shanti Shanti 🕉️
- ESP32-projects..
- ExecuTorch and Openvino for on-Edge inference..
- Accuracy-latency pareto curve in real-time detectors....
- Open-Vocabulary Detectors..
🔥 Come, sit in lotus pose, close eyes, feel the cosmic Wi-Fi… Good Vibes are flowing direct from Himalaya server!
Connect with me:

