2. “IT’S FANTASTIC TO SEE
THIS TYPE OF INNOVATION”
— Satya Nadella, CEO, Microsoft
3. 3
6.1
10.5
11.8
6.4
7.8
9.6
12.4
30.5
35.9
24.7
29.6
35.3
40.5
36.3
38.8
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
2012 2013 2014 2015* 2017 2018*
REVENUEINBILLIONU.S.DOLLARS
REVENUE IOT SUBSYSTEMS WW 2012 -2019 BILLIONS
CONNECTED CITIES
INDUSTRIAL INTERNET
WEARABLE SYSTEMS
CONNECTED VEHICLES
CONNECTED HOMES
“BUSINESSES ACROSS THE WORLD ARE RAPIDLY
LEVERAGING THE INTERNET-OF-THINGS TO CREATE
NEW NETWORKS OF PRODUCTS AND SERVICES, THAT
ARE OPENING UP NEW BUSINESS OPPORTUNITIES
AND CREATING NEW BUSINESS MODELS. THE
RESULTING TRANSFORMATION IS USHERING IN A
NEW ERA OF HOW COMPANIES RUN THEIR
OPERATIONS AND ENGAGE WITH CUSTOMERS.
HOWEVER, TAPPING INTO THE IOT IS ONLY PART OF
THE STORY. FOR COMPANIES TO REALIZE THE FULL
POTENTIAL OF IOT ENABLEMENT, THEY NEED TO
COMBINE IOT WITH RAPIDLY-ADVANCING ARTIFICIAL
INTELLIGENCE (AI) TECHNOLOGIES, WHICH ENABLE
‘SMART MACHINES’ TO SIMULATE INTELLIGENT
BEHAVIOUR AND MAKE WELL-INFORMED DECISIONS
WITH LITTLE OR NO HUMAN INTERVENTION.”
— PWC: Leveraging the Upcoming Disruptions from
AI and IOT, 2017
AI AND IOT
4. 4
“HOWEVER, DATA IS ONLY USEFUL IF IT IS
ACTIONABLE. AND TO MAKE DATA ACTIONABLE, IT
NEEDS TO BE SUPPLEMENTED WITH CONTEXT AND
CREATIVITY. IT IS ABOUT ‘CONNECTED
INTELLIGENCE’—WHICH IS WHERE AI AND SMART
MACHINES COME INTO THE EQUATION.
AI IMPACTS IOT SOLUTIONS IN TWO KEY DIMENSIONS
—FIRSTLY IN ENABLING REAL-TIME RESPONSES, AND
SECONDLY IN POST-EVENT PROCESSING, SUCH AS
SEEKING OUT PATTERNS IN DATA OVER TIME AND
RUNNING PREDICTIVE ANALYTICS.
THE INTERDEPENDENCE BETWEEN IOT AND AI ALSO
WORKS THE OTHER WAY. IOT’S CAPACITY TO ENABLE
REAL-TIME FEEDBACK IS CRITICAL TO ADAPTIVE
LEARNING SYSTEMS, SINCE OTHER TECHNOLOGIES
DO NOT REALLY ENABLE THIS ADVANCED TYPE OF AI/
ANALYTICS. SO THEY BOTH NEED EACH OTHER. THIS
COMBINED IMPACT OF IOT AND AI (ESPECIALLY DEEP
REINFORCEMENT LEARNING) WILL HAVE A DRAMATIC
IMPACT ON HOW BUSINESSES WILL TRANSFORM
THEMSELVES FROM A REACTIVE TO PREDICTIVE TO
AN ADAPTIVE BUSINESS MODEL.”
— PWC: Leveraging the Upcoming Disruptions from
AI and IOT, 2017
1073.7
1064.8
517.4
463.5
167.4 13.4 32.3
IOT UNITS INSTALLED BASE WITHIN SMART CITIES 2018
SMART HOME
SMART COMMERCIAL BLDG
TRANSPORT
UTILITIES
PUBLIC SVCS
HEALTHCARE
OTHER
AI AND IOT
5. “DEVELOPERS ARE MOSTLY FOCUSSED ON DELIVERING WELL-DEFINED FUNCTIONAL
REQUIREMENTS, AND BUSINESS MANAGERS ON BUSINESS METRICS AND
REGULATORY COMPLIANCE. CONCERNS AROUND ALGORITHMIC IMPACT TEND
ONLY TO GET ATTENTION WHEN ALGORITHMS FAIL OR HAVE A NEGATIVE IMPACT
ON THE BOTTOM LINE. BECAUSE AI SOFTWARE IS INHERENTLY MORE ADAPTIVE
THAN TRADITIONAL DECISIONMAKING ALGORITHMS, PROBLEMS CAN UNFOLD WITH
QUICKER AND GREATER IMPACT. EXPLAINABLE AI CAN FORGE THE LINK BETWEEN
NON-TECHNICAL EXECUTIVES AND DEVELOPERS, ALLOWING THE EFFECTIVE
TRANSMISSION OF TOP LEVEL STRATEGY TO JUNIOR DATA SCIENTISTS. INSUFFICIENT
GOVERNANCE AND QUALITY ASSURANCE AROUND THIS TECHNOLOGY IS
INHERENTLY UNETHICAL AND NEEDS TO BE ADDRESSED AT ALL LEVELS OF THE
ORGANISATION. WITHOUT XAI, GOVERNANCE IS VERY DIFFICULT.”
— PWC, Explainable AI, 2018
GDPR requires AI decisions involving personal data
to be explainable.
GENERAL DATA PROTECTION REGULATION
5
6. Envisioning Transformation
Our vision is to leverage the principles of human-
centered design and the power of eXplainable AI to
supercharge workforces and free humanity from the
current state tool- and app-based technology.
Design has become far too focused on the user
interface—tools, apps, human input and effort
comprising the user experience. Instead, we envision
technology that serves human objectives. We practice
the principles of human-centered design applied with
explainable artificial intelligence. Because we believe
most desirable technology experience is efficiency in
reaching the optimal end result.
HUMAN CENTERED DESIGN
6
7. 7
Our products are AI-powered agents that solve common
problems across specific job roles. Our agents are similar
to humans, except they look at the world a bit differently
– through a lens into the world of vast data.
Compared to the common AI, our assistants create a
closer, more trustful relationship with humans — a
bonafide advisor capable of answering questions and
making explanations in real world environments. XAI
brings humans and their digital assistants together on
task, fostering collaboration, on site, in the field, on the
sales floor and wherever workforce performance matters.
Deploy Senfino XAI at the network edge and fully
leverage the 4th industrial revolution!
HUMAN /MACHINE COLLABORATION
8. Explainable AI
At the heart of our technology is our
groundbreaking eXplainable artificial
intelligence (XAI). Leveraging decades of
advanced research, Senfino’s XAI engine
was built from the ground up in C++ and
employs a novel approach to neuro-fuzzy
and deep learning that provides for
interpretability and explainability in a
personalized, AI-assistant context.
XAI CORE:
y
1
y
N
1
1
1
y
2
X
1
X
2
X
n
y
α
β
.
.
.
.
.
.
.
.
.
.
.
.
Layer 1 Layer 2 Layer 3 Layer 4 Layer 5
AN
n MIN
A1
1
AN
1
A1
2
AN
2
A1
n
MIN
∑
α
β
∑
MIN
MIN
MIN
MAX
MAX
MAX
8
9. Explainable
XAI reveals the “why” behind complex data sets resulting in
recommendations, offering humans simply understandable
explanations to machine-based results. There’s no longer
any reason to take answers at face value. Explanations can
come in many forms, be it text, visual, audio or graph.
Interpretable
Not all explanations are created alike. Humans come with
various abilities to interpret information, so for example a
doctor would demand a different explanation on a
diagnosis than a patient, who speaks in lay terms.
Transparent
The ability see what’s happening inside a glass box can
have significant impact in curtailing algorithmic
discrimination and bias.
Auditable
In regulated industries with compliance scenarios, it’s
imperative to use XAI technology in automation. The ability
to audit processes and see how decisions are being made
can prevent significant legal exposure.
FROM BLACK BOX TO GLASS BOX
9
10. 10
FROM BIG DATA APPROACHES
TO PERSONALIZATION
The deeper your understanding of a problem, the better
prepared you are to solve it.
Big data offers tremendous opportunity to gain insights
from analytics, but the vast majority of data assets aren’t
harnessed or aren’t immediately actionable. Senfino XAI
offers a deeper understanding for more accurate insights
with a broader frame of reference. Senfino XAI analyzes
structured and unstructured data of various types such as;
real world 3D images, video, text, audio, financial and
market data, abstract topics, language patterns and
behaviors and IOT data. Whether that means individualized
recommendations, contextually aware personalized AI
assistance, just-in-time communications, mission critical
information or life critical process automation.
11. WE UNCOVER BEHAVIORS
We uncover the data logic behind the insights that can can be
captured in real world environments.
Value is created by shifting human behaviors in a company’s favor. To
accomplish behavioral change, AI must explain recommendation results of
machine learning systems in real world IOT environments.
11
“…WHEN MACHINES ARE ABLE TO
LEARN ENOUGH ABOUT THE
SITUATION AND MAKE RELIABLY
PREDICTABLE RECOMMENDATIONS
THAT HUMANS CAN TRUST, THEY
WILL BECOME AUTONOMOUS.”
— PWC
12. 12
We specialize in digital transformation strategy, design
and development of artificial intelligence for IOT applications
bridging machine learning with mobility through XAI agents via
smartphones, smart products, wearable tech and sensor, near-
field, edge, fog, narrowband and small cell networks.
We design and develop machine learning systems that deliver trainable
XAI assistants bringing value to the network edge.
INTELLIGENT IOT
“50% OF THE WORLD’S
TRADED SERVICES ARE
ALREADY DIGITIZED”
— McKinsey
14.
14
Our approach is highly iterative and tailored, inclusive and transparent
to ensure we identify AI solutions that generate value.
TRANSFORMATION PROCESS
Strategy
Design
Research
Design Development
15.
TRANSFORMATION KNOW-HOW
15
STRATEGY Value Planning
Requirements Engineering
EVO by Tom Glib
Value Proposition
Development
Service & Experience Design
User and Journey Mapping
DESIGN
RESEARCH
Context-Mapping
Shadowing
Mobile and Traditional Ethnos
In-person and In-Depth Interviews
Quantitative Segmentation
Conjoint and Discrete Choice
DESIGN Transformation Workshops
Design Sprinting
Rapid Prototyping
Machine Imaging
Neuro-Fuzzy Networks
Deep Machine Learning
DEVELOPMENT Back-End Systems & Cloud
Predictive Analytics
Prescriptive Recommender Systems
Process Automation
Chatbots, Alexa Skills, Google Actions
IOT and Smart Environments
AI is constantly evolving, we’re continually advancing.
Here are just a few of the ways in which we collaborate
with our client partners to discover, design, and realize
new futures.
TRANSFORMATION PROCESS
16. DATA ASSESSMENT RESEARCH INSIGHTS DATA STRUCTURE DESIGN & BUILD SUPPORT
SAMPLE APPROACH FOR PERSONALIZATION
• Perform technical analysis of
existing data warehouses and silos
• Develop initial ideas to consolidate
datastores for use in
personalization and XAI
• Conduct design research to gain a
rich grasp of needs, goals, and key
life moments, and how these
factors link to interactions with
communications.
• Merge human-centered and data
insights to create customer profiles
detailed enough to enable
personalized assistants
• Based on design research insights and
new consumer profiles, determine
what desired data is available but not
being collected
• Redesign data collection processes to
capture data identified as needed for
truly personalized actions
• Consolidate and unify datastores for
use in personalization & XAI
• Develop prototypes and estimate
business impacts and
implementation costs
• Analyze options and decide on
final personalization and XAI
strategy
• Build tools through a test, observe,
iterate approach.
• Evaluate outcomes against KPIs
• Set up regular check ins to review
performance and make any
updates
• Proactively identify any potential
updates not reflected in KPIs
• Detailed understanding of current
state data organization and types
of information included
• Initial solutions for transforming
datastore for personalization
and XAI
• New consumer profiles that
include detailed and nuanced
information that if collected can be
utilized to enable personalized
engagement
• Processes to collect new data
identified as useful
• Consolidated datastore for use in
BI, XAI, and/or personalization
• XAI-driven personalization
solution(s) designed to drive more
effective / efficient engagement of
consumers
• Iterative, insight-driven evolution
ACTIVITIES
DELIVERABLES
DISCOVERY DELIVERY ITERATION & SERVICE
16
17. SENFINO XAI BACKGROUNDER
The Senfino XAI Engine is engineered for human trust.
Senfino XAI employs a novel approach to machine learning.
Rather than leaving humans in the dark, Senfino XAI explains
the reasoning behind the given recommendation. Human
interpretable machine logic provides justifications to
recommendations concerning complex questions.
Senfino XAI engineering leverages decades of advanced
research by our own industry leading experts to optimize AI
processing of large disparate data sets, both structured and
unstructured, real world three dimensional images,
spectrographic images, video, audio, text, text analytics, IOT,
sensor and graph data.
The Senfino XAI Neuro-fuzzy architecture and Fast Computing
Framework leverages GPU processing for highly accurate results
at lightning fast speed.*
17
18.
SENFINO XAI RESEARCH
Senfino Content Based Recommendation System Using Neuro-
Fuzzy Approach provides human machine interpretable
explanation in an AI assistant context. Our Neuro-Fuzzy
architecture delivers substantial performance improvements
returning acutely accurate personalized content and
recommendations based on individual behavior without relying
on collaborative filtering (crowd sampling).*
Senfino Fast Computing Framework for Convolutional Neural
Networks (FCFCNN) embodies unique XAI architecture
reducing processing overhead while accelerating forward signal
flow. Neurons store reference pointers to corresponding regions
of previous input propagating signal flow, eliminating the need
to search for connections between layers. Additionally,
reference points are batched along with feature maps in multi-
feature input containers and treated as vectors, speeding
calculations across CNN layers. In benchmark tests of image
validation, FCFCNN performed twice as fast as the leading
OverFeat CNN.**
*Content Based Recommendation System Using NF Benchmarks.
**FCFCNN Benchmarks
18
123,281
121,809
127,206
126,921
126,423
126,804
127,093
126,011
0 50,000 100,000 150,000 200,000 250,000
GOLDFISH
GARTER SNAKE
TARANTULA
COONHOUND
COLOBUS MONKEY
DIGITAL CLOCK
FACE POWDER
GARBAGE TRUCK
FORWARD PROPAGATION TIMES FOR TEN EXAMPLE
CLASSES WITH VALIDATION IMAGES
SENFINO XAI OverFeat CNN
19. SENFINO XAI PERFORMS
Neuro-Fuzzy machine learning with real-time personalized XAI
assistants establishing trust through human machine experiences.
Optimized data assets, contextually relevant content, personalized
recommendations, behavior based micro-segmentation, demand
forecasting, GDPR, ECPA, FINRA, HIPPA, PII compliance.
Lightning Fast Convolutional Neural Networks analyzing large
volumes of unstructured data class and storage such as image,
audio, text, graph and semantic data in half the time.
Variational Autoencoding for reinforcement learning, probabilistic
modeling, probability matching and high velocity feed automation
from multiple databases and various data types.
Deep Autoencoding for statistical modeling of abstract topics
distributed across a collection of documents, systems and
databases.
19
20. INDUSTRY
Alliances and Channel Partners
Senfino has performed workshops, innovation initiatives and
hackathons through alliances with Microsoft, Ernst & Young and is
leading AI /IOT innovation with NXP.
Accolades
Our research on XAI is being published in world-renowned AI conferences:
2018 IEEE World Congress on Computational Intelligence WCCI 2018
The 17th International Conference on Artificial Intelligence and Soft
Computing ICAISC 2018
International Conference on AI and Soft Computing
Neural Networks (IJCNN)
International Conference on Parallel Processing
International Conference on Applied Mathematics
2015 IEEE International Conference
Big Data and Cloud Computing (BDCloud)
Social Computing and Networking (SocialCom)
Sustainable Computing and Communications (SustainCom)
ACM Interactions (a leading publication on Interaction Design)
20
23. LEADERSHIP
Officers and Advisors
Lukasz
Lesniak
Board
Member,
Senfino,
&
CEO,
Startberry
Tomasz
Rutkowski
CEO
/
Chairman
of
the
Board,
Senfino
MBA,
double
PhD
candidate
in
AI
and
InnovaFon
Management
Mark
Zurada
Co-‐CEO
/
Board
Member,
Senfino
Entrepreneur,
AJorney,
Engineer
Prof.
Leszek
Rutkowski
Professor
of
Computer
Science,
Member
of
Polish
Academy
of
Sciences;
President
of
the
Polish
Neural
Network
Society;
IEEE
Fellow
Prof.
Jacek
Zurada
Professor
of
Electrical
Engineering;
46th
most-‐cited
neural
networks
scholar
in
the
world;
nominee
for
IEEE
President
2020.
Member
of
Polish
Academy
of
Sciences
David
Harris
Investment
Banking,
Deutsche
Bank
Lech
Kaniuk
Former
CEO
of
iTaxi,
Poland’s
largest
taxi
hailing
app.
Former
CEO
of
PizzaPortal,
which
sold
to
Delivery
Hero
for
120M
PLN
Prof.
Danuta
Rutkowska
Professor
of
Computer
Science,
author
of
books
and
scienFfic
papers
on
ArFficial
&
ComputaFonal
Intelligence,
notably
arFficial
neural
networks,
fuzzy
systems,
geneFc
(evoluFonary)
algorithms.
23
24. ACCELERATOR
Our Warsaw-based AI accelerator and
community space backed by Microsoft and EY.
In concluding our first year of operation
recently, we accelerated 11 companies and
hosted 6,000 people at over 100 events.
Operates as a non-profit.
24
25. CASE HISTORIES
The common thread between the clients who we serve is simple:
we work with client partners that have embarked a journey of
digital transformation, realizing the importance of human centered
design and determined to harness the greatest value from IOT
systems with AI and machine learning.
25
26. Venture Picker is an AI-based assistant for the
Venture Capital community that provides accurate
and interpretable recommendations on investment
opportunities. Venture Picker uses Senfino’s novel
approach to neuro-fuzzy and deep learning. Venture
Picker will also be rolling-out experimental mixed
reality workstations, in addition to traditional web/
mobile apps.
26
27. As venture partners review and rank companies,
Venture Picker learns from each individual partner’s
unique behaviors.
As the machine learning system recommends
companies, the AI assistant offers a detailed
explanation as to why it has made each particular
recommendation.
27
We
know
that
the
last
equity
funding
amount
is
important.
In
this
case
it’s
$4,500,000.
It
seems
that
categories
is
important,
in
this
case
it’s
machine
learning.
28. We are currently collaborating with Optopol to develop advanced
medical image processing – supported by AI – and the design of a
system to help more quickly identify healthy vs. diseased cases.
The custom-build will be integrated into Optopol Technology's
machines system.
28
29. We are working in-concert with the University of Geneva to help
progress medical imaging and AI research in the area of detecting
cancer. The test is based on microscopic soft tissue samples, which are
analyzed on the basis of a biopsy; cell divisions that indicate tumor
outbreaks are counted, classified, and localized. The software aims to
facilitate the classification and determination of the stage of cancer
more quickly and accurately.
29
30. We worked closely with NYC EDC’s leadership to help envision a
new ‘home’ for their Futureworks initiative – a hardware accelerator
program based in New York City. We led value planning
workshops, conducted human-centric design research, and design
and developed the digital presence for Futureworks.
30
31. WORKSHOPS
Sample workshops for identifying opportunity /solution
hypothesis:
1. Discovery, stakeholder objectives, data and product value
opportunity
2. Business development, sales, marketing and customer discovery
3. Multi DB, Neuro-Fuzzy Machine Learning, IOT architecture
4. Candidate product innovation concepts, feasibility and impact
estimation
5. Actionable intelligence, workflow evolution and AI / XAI edge
network roadmap
31
“WITH SO MUCH AT
STAKE, DECISION TAKING
AI NEEDS TO BE ABLE TO
EXPLAIN ITSELF.”
— PWC
32. WORKSHOPS STARTING AT $25K
Discovery Workshop
Business needs and actionable insights via data science $25,000 — $50,000.
Three to Five Day Workshop $125,000 — $150,000.
Multi DB, Cloud /API, System and AI / XAI and IOT Network Integration Pending Discovery.
32
To learn more and schedule a workshop contact:
Barry Bryant, Senfino Partner /AI+IOT practice leader
barry@senfino.com or 917.651.1614