New Enterprises class at Sloan

Posted on the May 9th, 2007 under Business,Classwork,Sloan by

New Enterprises was a class I took at Sloan School of Management, taught by Professor Noubar Afeyan.  This class was on the subject of starting and running a business.  It had a lot of emphasis on venture capital, which is unsurprising because Professor Afeyan is one of the partners at Flagship Ventures.  Its other focus was on writing business plans.

I collaborated with a “Sloanie” (Justin Ashton) to develop a business plan for an online fashion magazine called Stylus.  It was not my top choice of a project either, but I was in it to learn. The magazine was a photo-heavy online affair with direct links to online stores where one could immediately purchase the clothes they liked.  We also wanted to create forums split along the lines of different subcultures of fashion, where stylists would hang out and dispense advice.

We had some ambitious plans for making it a tool for collecting demographics, conducting market research for new lines of clothes, and generating buzz for clothing lines from new designers. For the second phase, we planned to add in-house marketplace functionality to the magazine, where we would handle purchases, while leaving warehousing and shipping to the participating designer labels and factories. We have produced a thorough business plan and a flashy end-of-semester presentation. Feel free to take a peek at them below.

Stylus pitch

Stylus business plan

Financial Model

Financial model supplemental

Revenue Model

View Fullscreen

Modding Sonic Impact T-Amp Stereo Amplifier

Posted on the November 11th, 2005 under Hardware by

I modded a D-class digital amplifier (a Sonic Impact 5066) to improve the fidelity of low frequency sound. By adding two of beefy polypropylene capacitors in the preamp stage, I made the little power efficient amp sing like a top-notch, high-end amp worth many multiples of the original price ($40). I also replaced the original enclosure with something a bit more stylish, replaced the potentiometer by a vastly better one by Noble, and upped the DC power brick amperage to be able to drive my speakers better. Combined with two Titan bookshelf speakers by Paradigm, this setup sounds good.   To date, I was not able to find a high-end setup that was able to beat this combo’s performance so far in listening tests, no matter the price.  I think this has to do with the quality of the components and how well the amp characteristics match those of the speakers.

Out of the box, T-Amp is cheap and plastic-y looking

Out of the box, T-Amp is cheap and plastic-y looking

The rear panel of the unboxed T-Amp

This is the diagram of the so-called "stealth mod" by Audio1st

My modded T-Amp...ahh, much better. Love the industrial look of unpolished metal, and its contrast with the shiny knob.

A look inside. Note the large, high-quality polypropylene capacitors.

Intel Research at Berkeley

Posted on the May 13th, 2004 under Internships,Sensor Networks by

I had interned at Intel Research at Berkeley in the summer of 2004, under the supervision of Dr. Wei Hong. I started some work on sensor network monitoring which, later, has resulted in a paper of failure detection in wireless environments. I have also collaborated with a number of outstanding researchers, such as Joe Hellerstein, David Culler, David Gay, and many others, many of the original developers of TinyOS and subsequent developers at ArchRock . I have participated in the early stages of deploying a sensor network to monitor the Intel production facilities in Hillsboro, OR. We were also let into Intel’s top-secret D1D research plant, and remember watching the production line for what, I’m guessing, later became known as Core2Duo x86 processors, in total awe. I have finished the summer and received the highest commendation in my end-of-internship evaluation.

Sensor Network Programming

Posted on the May 13th, 2003 under Sensor Networks by

In order to program the sensor network, a group of grad students from NMS that was interested in sensor networks has created a fork of the Motelab codebase , with many extensions such as robustness to node failure and various hardware glitches, reprogramming retries, etc. My main contribution was a set of routines which allowed interaction with the Motelab web front-end from Perl scripts. I have also done a little work to help get EmStar to run on Mica2 motes, and integrate it with Motelab code.

Sensor Network Deployment

Posted on the March 13th, 2003 under Sensor Networks by

With a bit of help of other grad students from MIT Sensor Technologies group (MIST), I have deployed a 55-node wide-area sensor network testbed of Mica2 motes running TinyOS. I actually deployed this testbed three times, the first time in Tech Square 200 (which our building was still called LCS), the second time after our move to CSAIL, and the third time in order to equip our sensor nodes with enclosures. I also developed a toolset for programming the nodes and collecting experimental results.

A wireless sensor node in a custom-made enclosure tucked away on the shelf.

Sensors are exposed through the port in the sensor node enclosure.

The open case reveals the sensor node attached to the reprogramming board. This board was used to reprogram the nodes and collect experimental results over Ethernet. Of course, nodes in sensor networks "in the wild" would run on batteries and use wireless communication only.

Closeup: the Mica2 mote (left) and the sensor board (right) with light, temperature, and sound level sensors.