I'm not showing off. This is just what I am.

I'm an FDFS Freak — First Day, First Show. But in GenAI.

You know the type. The fan who books the earliest possible screening, front row, opening day. Not for the review scores, not for the crowd — for the sheer thrill …

UC Berkeley has open-sourced FreeToken, an edge-native inference engine that runs frontier-scale Mixture-of-Experts (MoE) models on consumer hardware. It treats GPU, CPU, host memory, and PCIe interconnects as a single unified inference platform.

The approach works because of how MoE models are structured. A model like DeepSeek-V4-Flash carries 284B total …

September 5th is Teachers' Day, marking the birthday of Dr. Sarvepalli Radhakrishnan — philosopher, statesman, and India's second President. He once described casteism as a cancer afflicting India. In that spirit, this piece revisits his views on the subject and asks that we treat Teachers' Day as an occasion to confront …

A claim from social researchers in Tamil Nadu holds that "Thevar" was never a caste name at its root — it was a title given to the leader of a rural place, and only later came to be read as a community label. There is significant truth in this. It reflects …

Every Kiro spec opens with a small but consequential fork: what do you author first? The IDE offers two starting points — Requirements or Technical Design — and treats Requirements as the recommended default. The choice isn't cosmetic. It decides which artifact is the source of truth and which is derived, and …

A toponymic surname is a family name derived from a place — a village, town, city, region, or geographic feature. It is one of the oldest and most widespread ways human families have come to be named, and among Telugu people it is close to the rule rather than the exception …

WeeLLM is a layer-streaming inference engine for large diffusion models. Its one claim is unusual: run a ~12B FLUX model, or even a ~20B Qwen-Image model, in under 4GB of VRAM — with no quantization. Full bfloat16 weights, degraded only in speed, never in precision.

The Core Idea

Layer Streaming The …

The frontier of AI is no longer just about who has the biggest model. It's about who can run intelligence cheaply, quickly, and everywhere. That shift is what Lean Inference is about: treating inference cost, latency, and accessibility as first-class engineering problems rather than afterthoughts.

For years the industry chased …

Mark Zuckerberg went from building a home messaging app at age 12 to running the world's largest social network. This is the story of how a self-taught programming prodigy turned a dorm-room project into a platform used by billions — and the controversies he weathered along the way.

The Kid From …

Everyone has their dark days. Here are the toughest situations I went through — the ones that shaped my life and taught me to face any challenge that comes my way.

The Fall

Second Class Graduate — I graduated second class in Engineering, and no employer would even consider me for an …