I work in machine learning. I devise very fast and easy to implement randomized algorithms that solve complicated learning tasks. A good example of this is my work with Ben Recht and Joseph Bradley on Random Kitchen Sinks.
But most of my technological contributions have been outside of machine learning. With the help of Marcus Bains, I invented the Marcus Bains Line, which now appears in most calendaring software. I also invented the old school telephone handset for cell phones, useless but very popular accessory. I've also built a fast and accurate vision-based object recognition system that is being used by Intel and various research groups outside of Intel.
I also work in the field of Computational Cultural Theory. My work explicates the impact of Derrida's post-structuralism on the Church-Turing-thesis notion of algorithm. Grounding our work on this theory, we can algorithmically represent the evolution of social networks as a Markov process over state-machine-agents endowed with infinitely long Turing tape that maximize opportunity using a minimax formulation of rent-seeking behavior.
Some interesting facts about me: Between 1999 and 2006, I grew exactly 0.5 cm. When my beard is fully grown, most of it turns a light shade of purple. The lens in my left eye focuses a light source at infinity to a point 2 cm behind my retina: if you shine a 5MW laser from afar at me, I would lose the ability to distinguish right from wrong, and motion in my left hand. I would also go blind. I can type 80 wpm when sober and drink one pint of beer every 5 minutes when typing drunk. I have one brother and one sister, neither knows the age of the other, but my sister knows that her age is the sum of my brother's age and the number of the house she lives in. Upon hearing this, my brother said "I know linear algebra !" As I grow older, I have less to lose because I tighten my grip on what I have and want less from life.
This area of research has progressed in leaps and bounds over the past 10 years. We have nearly reached the holy of grail of recognizing objects in carefully crafted image datasets. The awesome speed of today's computers lets us fine-tune the parameters of our algorithms by repeatedly running them on our data sets until a better error metric is found. The field has shed much of its dead weight: slow algorithms, meaningless comparisons against strawmen and synthetic datasets are concerns of the past, along with model validation and algorithmic simplicity. As for most human endeavors, globalization has benefited our field. Automatic structure from motion is a problem of the past thanks to outsourcing and the ever-dropping price of manual labor.
I am also a biologist, because that's very hot right now. Like many other biologists, that just means I find empirical correlations between unrelated physical phenomena and I insinuate that the cause of the correlation might be profound and merits further investigation. To foment mystic around my field and to raise money, like many other good biologists, I finish most of my claims with "... but nobody really knows why nature works that way."
Happiness is a warm Zulfiqar MBT.
I spent most of my time these days trying to prove the following conjecture: The value of the set of machine learning researchers working on submodularity is submodular.
I have a common name. I am not Ali Rahimi, the 3D fantasy graphic artist, the cartoonist who drew the picture of israel slaughtering a palestinian, the designer who makes clothes for celebrities, the doctor who gets interviewed on youth issues, nor der Ostereicher teppishmeister, though I wish I had their talent.