Commercilising Grids, and accademics view

1
Commercialising Grids, an academics view Jim
Austin
2
Overview

• My position
• Cybula who are we?
• Development of our Grid solution
• DAME
• BROADEN
• CARMEN
• Lessons learned

3
Academic or industrialist

• I am
• CEO of Cybula and Head of Research group
• Gives a unique view from both sides
• Slightly unusual York allows this
• In my view its ideal

4
Cybula

• Founded 2000
• Spin-out of University of York, Computer Science
• Privately funded and owned company
• Based in IT centre at York Science Park
• Draws on 35 staff

Cybula
5
Cybula

• Subcontracts between the University and the
  company both ways
• Very flexible for resources, skills
• University offers
• Skilled people, access to technology, access to
  government funding
• Cybula offers
• Product development, market view, support
  services, development, commercial funds

6
Company Aim

• Sells high performance pattern matching software,
  hardware and skills
• Provides RD to support customers needs
• Works with end users in vertical market segments
• Started with project focus now adding products
  as they mature in the projects

7
Markets

• Any market where managing large complex
  unstructured data is a problem
• Legal
• Biometrics
• Pharmaceuticals
• Engineering

8
Technology
9

• AURA Advanced Uncertain Reasoning Architecture
• Set of neural network based methods for dealing
  with complex data

10
Advantages

• Depend on application area
• Fast and scalable for complex data
• Other methods built on conventional database
  methods slow for this type of data

11
(No Transcript)
12
Products

• Main products
• Signal Data Explorer Grid based signal search
  and analysis tool
• FaceEnforce 3D Facial Biometric system

13
Grids

• Cybula developed an understanding of Grids in
  partner ship with University
• Initially just watching
• Then involved in trials
• Now develop products
• Careful, relatively low risk, approach

14
What did it offer?

• Cybula had a local search engine, but not an
  enterprise solution
• Grids offered the possibility of
• Distributed search and analysis of data
• Allowing users access to data any where, without
  knowledge of location

15
Grids

• Cybula involved in a number of EPSRC and DTI Grid
  projects since 2000
• DAME 2000-2003 associated partner
• BROADEN 2004-2007 full partner
• CARMEN 2007-2010 exploitation partner
• All used AURA search technology within the
  application

16
Development

• DAME
• This allowed University of York to test the ideas
  learn the technology painful!
• Technology developing
• No one really understood it
• Partners were try to develop a good business case

17
Data focus

• Aimed at the search and analysis of signal data
• Very large amounts of complex data
• Find-one-like-it approach

18
Development

• At the end of DAME all was clear we had
• A good piece of technology Signal Data Explorer
  a GUI for Vibration data search and analysis
• A distributed search engine
• G-AURA pattern match module
• Grid technology for managing distributed search
• Pattern Match Control (PMC)

19
PMC
GetResults()
ReturnResults()
Search()
20
SDE GUI
Location of master node
Spectral data
Search results
21
Evaluation

• Cybula then took the plunge
• Involved in BROADEN, DTI funded
• Also with Rolls-Royce
• As the exploitation partner
• Helped define the product
• Looked at possible markets

22
CARMEN

• A new project CARMEN allowed us to define a good
  target market
• Neuroscience
• Simple choice ready and willing collaborative
  team
• Area we knew well

23
Some Customers

• Bombardier use SDE for improving train
  diagnostics and prognostics (made Virgin trains,
  under ground trains etc.)
• Vodafone/Sun use PMC to manage image search
  match problems banned images

24
Lessons

• It takes time
• Building a relationship with a company is
  essential
• Trust and understanding on both sides is
  essential
• Easy with Cybula, and independent company could
  be more difficult

25
Lessons

• General Grid capability is very valuable
• Implementation and experience of large grid
  systems will take you further than others
• This comes off the back of UK eScience initiative

26
Lessons

• We did not commercialise early enough
• I was waiting for Roll-Royce validation of the
  technology
• They are a great early adopter
• Big companies take ages (years) to do this
• Can tackle other markets in the mean time

27
Lessons

• Grids on there own dont hack it (for us)
• Grids enabling/enhancing an existing technology
  do
• Technology is always changing a pain, but also
  an opportunity

28
Lessons

• You dont introduce more than one new innovation
  at a time, i.e.
• We had AURA introduced Grids for distributed
  search
• We would not introduce AURA and Grids together
  too much risk for a customer two untried
  technologies.
• Leverage is weaker, can manage the market better
  in a stepwise approach

29
Lessons

• Collaboration with Universities is essential in a
  new market
• Too risky otherwise
• Good access to early adopters
• Access to skills
• Access to expensive - new technology
• (that would be my view!)

30
Thanks!

• www.cybula.com
• (44)1904 56 76 86