28
Henri Blondelle www.agileDD.com March 2017 What ever is your business, structured information delivers reliable decisions

R=233 - G=229 - B=223 What ever is your business ... · Chemical, Automotive, Aeronautics •Our Machine Learning System is not dependent on particular datatypes, only ... © 2017

  • Upload
    dothuy

  • View
    215

  • Download
    2

Embed Size (px)

Citation preview

Henri Blondelle www.agileDD.com

March 2017

R=233 - G=229 - B=223

R= 33 - G= 39 - B= 69

R= 78 - G=103 - B= 200

R= 94 - G= 204 - B= 243

R=167 - G=234 - B= 82

R=149 - G= 98 - B= 81

R=145 - G=132 - B=133

R=241 - G= 65 - B= 36

What ever is your business, structured information delivers

reliable decisions

50% of the Fortune 500 disappeared in the last 15 years

© 2017 Agile Data Decisions – Confidential 2

and a lot of others have been ... successful

© 2017 Agile Data Decisions – Confidential 3

the reason why?

© 2017 Agile Data Decisions – Confidential 4

more data more metadata

more indexed data

more sourced data

more verified data

more success

© 2017 Agile Data Decisions – Confidential 5

But, more data is not always new data

Legacy document

assets are rich in

metadata ready to

be structured in DBs,

But indexing and

cataloging is

expensive and time

consuming

© 2017 Agile Data Decisions – Confidential 6

Our ambition is to automate the indexing process

© 2017 Agile Data Decisions – Confidential 7

And to structure 80% of the unstructured

information, reversing the current proportion*

20%

80%

*Sources: Merrill Lynch; EMC; GATE project/Sheffield University studies show that in average, 80% of the available information is unstructured

© 2017 Agile Data Decisions – Confidential 8

SAVE

MONEY

Avoid populating databases manually

GO

FASTER

From data to decision

DE-

RISK

Using more verified information

Creating value from dormant data assets R=233 - G=229 - B=223

R= 33 - G= 39 - B= 69

R= 78 - G=103 - B= 200

R= 94 - G= 204 - B= 243

R=167 - G=234 - B= 82

R=149 - G= 98 - B= 81

R=145 - G=132 - B=133

R=241 - G= 65 - B= 36

© 2017 Agile Data Decisions – Confidential 9

iQC Machine Learning detects metadata …

and more!

© 2017 Agile Data Decisions – Confidential 10

Learning

models

Structured

information

publication

QC

Data edition

Validation

Refutation

text

searchable

OCR

Pre-OCR

processing

Post-OCR

processing

Parallelized OCR

User provided

taxonomy

(option)

Seeds docs

(option) Seeds db

(option)

Heuristic labelling (option)

iQC Machine Learning

Documents classification

Metadata extraction

iQC

datalake

GUI

Unstructured and

semi-structrured

documents

A uniform

text layout

representation

How it works?

© 2017 Agile Data Decisions – Confidential 11

iQC environment

© 2017 Agile Data Decisions – Confidential 12

iQC v1.0 has been successfully deployed

6 man - years of development

Patent claim registered Dec 2016

Existing Customer

Published Work (SEG,

PNEC)

© 2017 Agile Data Decisions – Confidential 13

Most promising IT&web startup

2017 by RUA

Joe Johnston, Head Petrophysicist, CGG: “iQC is not only a way to reduce the cost of populating data-bases, it is a way to to source and QC massive information before any analysis”

Mike Davidson, Guest editor of The Leading Edge – March 2017, SEG, “A considerable challenge faced by many working geoscientists is the time spent searching for, wrangling with, and making sense of large legacy databases, many of which may lack a natural organizational structure. This unstructured data may include such information as field notes from ops-geologists on vintage mud-logs — potentially valuable if automated methods could be employed to make sense of it. AgileDD shows how machine learning could be employed in the management of databases.”

Feb. 5th 2017 © 2017 Agile Data Decisions LLC - Strictly Confidential

Strong and increasing interest from the industry

15

Manual indexing

OCR + Full text indexing

OCR + Full text indexing + Machine learning

Cannot process a lot of documents in a short period of time

Can automatically extract metadata already known in lists, dictionaries, taxonomies ...

Because the ML searches for context around the metadata, text and numerical variables can be detected

Ahead from competition

And ready to access a $100M+/year market in the E&P industry

• Average IOC spends $10M/year to catalogue. Total expenditure around $0.5B

• Millions files are stored with very few associated metadata

• Cataloging has to be redone several times

• Issue affects all the players in the value chain

© 2017 Agile Data Decisions – Confidential 16

Gain traction in E&P sector first

• E&P industry is data dependent.

• It’s a market segment where we know the data, the data environment,

data consumers and their expectations

• IOCs, NOCs, Independents, Unconventional, Regulators, Contractors ...

All have the same unstructured data issue

• Our product benefits from our launch partner’s expertize and user experience

• After a period of adaptation due to oil price decline, O&G sector is now looking at emerging technologies to perform in the new O&G environment

Automation is the only way to effectively make use of the unstructured data in the E&P market

© 2017 Agile Data Decisions – Confidential 17

O&G is a $100M+ market segment. We target 5% in the next 3 years.

© 2017 Agile Data Decisions – Confidential 18

SaaS+ business model

© 2017 Agile Data Decisions – Confidential 19

reve

nu

e

cost

Nb customers

Lice

nce

+ m

ain

t +

R&

D

Hyb

rid

+

con

sult

ing

SaaS + subscription SaaS on demand

iQC for free

• Customer pay per document and extracted information

• In line with the cloud platform, internal or external

• Pricing aligned on value creation

• OPEX cost only, easy to be compared with the manual process cost

• Collaborative learning models, we capitalize on customers experience

• Open a market place for structured data exchange

• Collaborative learning models, we capitalize on customers experience

• Open a market place for structured data exchange

Extension to other domains and industries

• Indexing and cataloging unstructured documents is a common issue in all O&G domains: development, production, asset integrity, downstream and all industries: Chemical, Automotive, Aeronautics

• Our Machine Learning System is not dependent on particular datatypes, only our Learning Models are

• iQC Graphic User Interface can be adapted to other document types and SMEs needs

OCR ML GUI

docs models

results iQC Reusable elements for a pivot

© 2017 Agile Data Decisions – Confidential 20

• The first 15 months

• Focus on exploration documents (wells and seismic related), establish a revenue stream

• Make 2 or 3 permanent reference clients

• Strengthen the software development team to add new features requested by early adopters

• Build partnerships with cloud resource vendors

• Re-use the iQC platform to create an MVP allowing AgileDD to open a new industrial market segment

• Prepare the level-A funding

Post Level-A agenda

• Increase our share O&G market enlarging the iQC technical

scope

• Penetrate the second market

• Hire sales and product staff knowledgeable of that market

• Agile development based on the existing MVP

• Develop new partnerships

• Incremental expansion on a new market segment

AgileDD agenda

© 2017 Agile Data Decisions – Confidential 21

Sales projection

© 2017 Agile Data Decisions – Confidential 22

$0

$2,000,000

$4,000,000

$6,000,000

$8,000,000

$10,000,000

$12,000,000

2016 2017 2018 2019 2020

Sales per service - 5 Years Forecast

Energy Sector Insurance/Pharmaceutical Engineering and Manufacturing Consulting & Integration Services - Maintenance

Cash flow projection

© 2017 Agile Data Decisions – Confidential 23

2017

Today

May Jul Sep Nov 2017 Mar May Jul Sep

Agile Data Decisions LLC Creation

5/26/2016

AgileDD website v1.0

7/6/2016

iQC @ SPE/IE Aberdeen

9/7/2016

CDA results @ ECIM - Stavanger

9/13/2016

iQC @ Saudi Aramco - KSA

11/2/2016

AgileDD website v2.0 and linkedin page

11/15/2016

iQC at PETEX with CGG - London

11/16/2016

CDA results @ CDA Aberdeen

11/30/2016

iQC @ Shell North Sea - Aberdeen

12/1/2016

iQC @ IFP - Paris

1/9/2017

AgileDD poster at EVOLEN - Paris 1/16/2017

RUA Most Promizing Startup - Houston

1/19/2017

iQC @ TOTAL - Paris

1/25/2017

2200 views on AgileDD linkedin page

2/2/2017

iQC @ Anadarko - Houston

2/9/2017 iQC @ Atos-Bull - Prais

2/18/2017

TLE long paper about iQC

3/6/2017

iQC @ TOTAL Norge - Stavanger

3/7/2017

CDA results ECIM - Stavanger

3/8/2017

iQC @ Statoil - Stavanger 3/9/2017

iQC @ OMV - Vienna

3/15/2017

AgileDD speed dating @ Avenia - Pau

3/16/2017

AgileDD @ AAPG Pitch Appalooza

4/13/2017

AgileDD @ RUA-OTC

5/1/2017

AgileDD @ PNEC - Houston

5/16/2017

AgileDD @ Energistics NDR - Stavanger

6/6/2017

EAGE ML workshop (submitted)

6/12/2017

AgileDD @ ECIM 2017 - Stavanger

9/12/2017

Marketing

Opportunities by stage

Deal

Submitted quote

Analyzed need

Presentation and demo

Contacted for demo

Events and Social Media

Founders

Dr. Amit Juneja CEO & CTO 16 years of experience in Data Science and ML

Jacques Micaelli President 20 years of experience in IT

Henri Blondelle Sales and Marketing 25+ years of experience in O&G data management

Philip Neri Marketing & Investor relations

Joe Johnston Geosciences

Strong Team with complementary talents

Independent advisors

© 2017 Agile Data Decisions – Confidential 26

• Using ML to make unstructured data directly accessible

• Targeting 5% of the $100M+ primary E&P market

• Solution can be easily adapted to other industries

• Solution tested by a major geoscience service provider

• Established and competent team

• Raising $750k pre-A to develop E&P market and enter in a second market

Takeaways

© 2017 Agile Data Decisions – Confidential 27