self 
    I was born in Madras in India (now called Chennai). I lived in Madras and Delhi until I was five years old. My father, who worked for the Indian Government was posted in Washinton DC in 1967. We lived there for a few years and attended Janney elementary school.

St Mary's school, Mazgaon, Mumbai    After returning to India, I lived in Bombay (now Mumbai) for 5 years where I attended the ISC section of St. Mary's school (picture on the right) . The best thing about this school was that the Central Railway tracks ran by the school.  I used to spend every lunch break near the fence watching the trains.  Music class (Mrs. Saldanha) was especially difficult since it was right next to the tracks. Instead of paying attention to the music, I used to spend my time looking out of the window!

    We moved to Delhi in 1975 and finished high-school at the (take a deep breath ) Bharatiya Vidya Bhavan's Vallabhram Mehta Public School.

    I got through the JEE (Joint Entrance Examination) and studied at Indian Institute of Technology, Delhi (picture below) from where I graduated with a B-Tech in Textile Chemistry in 1984. I stayed in Nilgiri Hostel. My IIT experience was like most - intensive study with a lot of fun thrown in between.
IIT Delhi   
    Despite my odd choice of major, I actually managed to take almost every course offered on computing at that time. Given the frenzy over IT in India nowadays, its hard to believe how sleepy things were back in the early 80s. We had one big ICL 2960 mainframe on which I learned programming, most of which was in Fortran. The system had a primitive editor in which one could only make one pass through a file - no way to go backwards! Even so, the computing facilities were far superior to most other universities in India at that time.

       My B.Tech project was on computer color matching and I'm still amazed that I got it to work. The idea is to be able to take three dyes in the primary colors and obtain the concentrations to use to obtain any desired color. It was my largest program to date - an optimization algorithm implemented in a couple thousand lines of Fortran.

    Like many IIT graduates at that time, I wrote the GRE and went to the USA study computer science at Syracuse University.  I started research in parallel programming but after a couple of years, did not seem to be getting anywhere. I then switched to the field of neural networks. I got the idea for my dissertation (modular neural networks) when I was playing with a complete backpropagation network that I had programmed in a spreadsheet (sc if I recall correctly). I noticed that the signs of the weight changes were different depending on the exemplars in the training set - this was for a classification problem. From that small observation, I was able to spin out three journal papers and was able to complete my dissertation - a good thing since I was getting a little anxious to graduate! I did finally graduate with a Ph.D. in Computer Science in 1992.

    In 1991, I joined the IBM T. J. Watson research center in Hawthorne, NY as a Research Staff Member in the networking group (nothing to do with neural networks whatsoever!) I worked on a number of projects in this group including link-layer software for managing an early optical network called Planet/Orbit and a network test infrastructure.

    The Orbit network was a high-speed optical ring network with ATM style header processing but with provision for variable size payloads. Due to a bug in the TriQuint Hot Rod chip, there was no easy way to bring up the network. Each station depended on the upstream station to help it to help it become active. I came up with a distributed algorithm to solve this problem. I used to joke  that TriQuint was responsible for keeping me employed in the early 90s.

    It also became clear that testing network equipment had some special challenges. In order to create a reproducible test case, many parallel activities need to be coordinated. To solve this, I came up with a parallel implementation of Tcl that I called Andes. From a central control station, I used to send out TCL programs to be executed by agents. The scripts would all be executed and the results collected on the central station.

    After a few years of working on low-level networks, I decided to move up the network stack a bit. From 1996 to 1998, I worked on  a variety of projects including system administration tools and a capability system for Java applets (patent 6,044,466). I then moved to the E-commerce group and had a great time there for two years.  I worked on auction and reverse auction systems as well as an e-coupon distribution system.  I was the principal designer for a web-based procurement system for IBM that was actually put into production.  I also worked with IBM global services on a number of interesting projects.

    I joined Reefedge Networks in 2000 as a founding employee. The main product is a wireless LAN security system. In 2001, I became the director of applications and middleware.  In other words, I managed most of the product functionality that was not within the kernel. I have worked on many aspects of the product including system administration, monitoring and configuration management.  Some of the work is covered by patent 6,691,227.

    In September 2004, I returned to the IBM T. J. Watson Research Center as a reseach staff member in the e-commerce business intelligence group headed by Doug Riecken. I am currently working on some interesting applications of AI techniques to business problems.

Copyright © Rangachari Anand 2004