● Hyderabad
Praveen Chandrahas
I like to build things, think about how intelligence and energy / entropy relate (on a cosmic scale - more below), and tinker with stuff.
I have spent most of my working life moving between engineering, running teams, and helping start organisations. Much of that has been in and around the free software movement in Telugu-speaking regions, and in building technology that works in local languages and for working people.
I am more at home with systems, electronics and integration than with writing elegant code. I like understanding how a thing actually works, and checking that it does. I live here with my family and a beagle named Max.
01 Work
Two things occupy most of my working time at the moment, both built with co-founders.
Inference · infrastructure
Making AI inference tractable
Work on making AI inference cheaper and more tractable to run on infrastructure an organisation controls.
Education · skilling
Applied AI skilling
An AI-skilling effort aimed at practitioners and faculty who need to use these tools well, not just read about them.
Before these, I co-founded an enterprise AI company in 2021 and have since left it.
02 Research
On the side, and out of curiosity, I work on one question about entropy, life and agency:
We know the answer. We are proof that it does arise. So, the question is, if we set similar conditions, would agency arise in machines? Can AGI / AI / emergent properties come out of a similar set of conditions? That's the question I keep pondering over. What are the initial conditions? And what ultimately led to life and intelligence as we know it? If we 'program' those conditions and constraints, would machines become sentient? Would agency arise?
Paper · speculative
Why AGI is a thermodynamic problem
An argument that the limits on intelligence may be better framed in terms of sustained entropy production than in terms of algorithms alone. It is an argument, not a result, and I treat it that way.
Code · preregistered
agency-from-gradients
A public simulation suite that tests the question above in small, simplified worlds. The predictions were written down before the runs, so a failed prediction counts as much as a successful one.
03 Tinkering
These are hobbies. I do them for the satisfaction of understanding each part, not to chase a particular outcome.
Audio · DSP
A software-defined home theater
Building the audio pipeline stage by stage and measuring each one. The interest is in verifying what each component does, not in audiophile claims.
Code and notes →Optics · sky
A DIY planetarium and a smart telescope
Partly for myself and partly to share the night sky with family.
Outdoors
Dark-sky trips
The Western Ghats and the forests of Telangana are my favourite places to go and look up.
04 Musings
Occasional essays.
Essay
Curiosity killed the cat. So we killed curiosity.
Debunking can't keep up with the claims. A case for playing the long game instead: keep science, and the wonder of it, alive in the open.
Read the essay →