Sentinel Week 4 H4D Stanford 2016

agile, bmnt, business model, corporate innovation, customer development,dod,nsa, socomm, diux, h4d, hacking for defense, lean, lean launchpad, lean startup, stanford, steve blank

1. Team Sentinel Team members: Jared Dunnmon Darren Hau Atsu Kobashi Rachel Moore Cumulative # of interviews: 39 + 13 Users: 4 Experts: 9 What we do: Enable rapid, well-informed decisions by establishing a common maritime picture from heterogeneous data Open and automated data aggregation (i.e. incorporate open source data) Flexible layering and filtering with improved UI/UX Enhanced intel through contextualization and easily accessible, common database Identifying deviations from baseline Why it matters: Information overload A2/AD prevents deployment of traditional ISR in a timely manner Data aggregation platforms and database access in PACOM appear extremely manual Military Liaisons John Chu (Colonel, US Army) Todd Cimicata (Commander, US Navy) Problem Sponsor Jason Knudson (Lieutenant, US Navy 7th Fleet) Tech Mentors include: Elston ToChip (Palantir)

2. Customer Discovery Hypotheses Experiments Results Action We want automated data layering REFINED - Interviews with CMDR Pablo Breuer (N3/6), CMDR Rob Williams (N0) - There is too much layering going on - Info overload, esp in Straits of Malacca - Need a better filtering method - Revamp the MVP to focus on filtering layers - Get a sanitized GCCS screen of Straits of Malacca A common, easily-searchable database is desireable VALIDATED - Interviews with CMDR Rob Williams (N0), CMDR Chris Adams (N3), CTR3 Joseph Baba (N2), CMDR Pablo Breuer (N3/6) - Engagement with Elston ToChip (Palantir) - Interview with Chad Dalton, Pat Kelly (OGSystems) - Need to submit help ticket to access certain databases - Palantir often interacts with customers who have siloed datasets - 6+ databases that people search through manually - takes hours - Develop separate MVP to test database functionality - How to establish common database without losing context-specific attributes? - What are the data feeds for 7th Fleet vs. PacFlt vs. PACOM? Analysts will analyze, no need to do our own algorithms INVALIDATED - Interviews with Dr. Evelyn Dahm (SOCPAC advisor, J5/8), CMDR Rob Williams (N0), CMDR Silas Ahn (N2) - Would be helpful to establish baseline and alert when anomaly occurs - Want push notifications when a ship violates known patterns - Determine example scenarios that 7th Fleet wants to monitor

3. Customer Discovery Hypotheses Experiments Results Action We want automated data layering REFINED - Interviews with CMDR Pablo Breuer (N3/6), CMDR Rob Williams (N0) - There is too much layering going on - Info overload, esp in Straits of Malacca - Need a better filtering method - Revamp the MVP to focus on filtering layers - Get a sanitized GCCS screen of Straits of Malacca A common, easily-searchable database is desireable VALIDATED - Interviews with CMDR Rob Williams (N0), CMDR Chris Adams (N3), CTR3 Joseph Baba (N2), CMDR Pablo Breuer (N3/6) - Engagement with Elston ToChip (Palantir) - Interview with Chad Dalton, Pat Kelly (OGSystems) - Need to submit help ticket to access certain databases - Palantir often interacts with customers who have siloed datasets - Analysts are spending too much time developing their own custom ETL tools - Develop separate MVP to test database functionality - How to establish common database without losing context-specific attributes? - What are the data feeds for 7th Fleet vs. PacFlt vs. PACOM? Analysts will analyze, no need to do our own algorithms INVALIDATED - Interviews with Dr. Evelyn Dahm (SOCPAC advisor, J5/8), CMDR Rob Williams (N0), CMDR Silas Ahn (N2) - Would be helpful to establish baseline and alert when anomaly occurs - Want push notifications when a ship violates known patterns - Determine example scenarios that 7th Fleet wants to monitor

4. Customer Discovery Hypotheses Experiments Results Action We want automated data layering REFINED - Interviews with CMDR Pablo Breuer (N3/6), CMDR Rob Williams (N0) - There is too much layering going on - Info overload, esp in Straits of Malacca - Need a better filtering method - Revamp the MVP to focus on filtering layers - Get a sanitized GCCS screen of Straits of Malacca A common, easily-searchable database is desireable VALIDATED - Interviews with CMDR Rob Williams (N0), CMDR Chris Adams (N3), CTR3 Joseph Baba (N2), CMDR Pablo Breuer (N3/6) - Engagement with Elston ToChip (Palantir) - Interview with Chad Dalton, Pat Kelly (OGSystems) - Need to submit help ticket to access certain databases - Palantir often interacts with customers who have siloed datasets - People use data analysis products to support preconceived ideas - Develop separate MVP to test database functionality - How to establish common database without losing context-specific attributes? - What are the data feeds for 7th Fleet vs. PacFlt vs. PACOM? Analysts will analyze, no need to do our own algorithms INVALIDATED - Interviews with Dr. Evelyn Dahm (SOCPAC advisor, J5/8), CMDR Rob Williams (N0), CMDR Silas Ahn (N2) - Would be helpful to establish baseline and alert when anomaly occurs - Want push notifications when a ship violates known patterns - Determine example scenarios that 7th Fleet wants to monitor

5. Customer Discovery Hypotheses Experiments Results Action We want automated data layering REFINED - Interviews with CMDR Pablo Breuer (N3/6), CMDR Rob Williams (N0) - There is too much layering going on - Info overload, esp in Straits of Malacca - Need a better filtering method - Revamp the MVP to focus on filtering layers - Get a sanitized GCCS screen of Straits of Malacca A common, easily-searchable database is desireable VALIDATED - Interviews with CMDR Rob Williams (N0), CMDR Chris Adams (N3), CTR3 Joseph Baba (N2), CMDR Pablo Breuer (N3/6) - Engagement with Elston ToChip (Palantir) - Interview with Chad Dalton, Pat Kelly (OGSystems) - Need to submit help ticket to access certain databases - Palantir often interacts with customers who have siloed datasets - JIOC, PacFlt, 7th Fleet do not see the same feeds -> may lag each other by 2-6 hours! - Develop separate MVP to test database functionality - How to establish common database without losing context-specific attributes? - What are the data feeds for 7th Fleet vs. PacFlt vs. PACOM? Analysts will analyze, no need to do our own algorithms INVALIDATED - Interviews with Dr. Evelyn Dahm (SOCPAC advisor, J5/8), CMDR Rob Williams (N0), CMDR Silas Ahn (N2) - Would be helpful to establish baseline and alert when anomaly occurs - Want push notifications when a ship violates known patterns - Determine example scenarios that 7th Fleet wants to monitor

6. Coast Guard data fusion center visit scheduled for Friday (4/22) Customer Discovery

7. Research - Interviews to assess needs, organizational dynamics, procurement strategy - Site visits to see current practices -Understanding current workflow Prototype - Evaluate existing sensor platforms with commercial partner s - Integrate sensor feeds of interest into prototype platform - Compile existing data resources - Create representative “fake” datasets - Evaluate relevant ML algorithms for prediction and rules for push alerts - Iterate on human-machine interaction Strategic Decision Makers VADM Joseph Aucoin ADM Scott Swift (PacFleet) ADM Harry Harris (PACOM) Analysts (N/J2) E.g. Jason Knudson, John Chu, Jed Raskie, Joseph Baba Operators (N/J3) CDR Chris Adams (7th Fleet) Planners (N/J5) Need to find these people - Common and consistent view of the Area of Responsibility (AOR) - Timely operational decisions - Decreased time to predict hot spots, ID & differentiate threats - Reduced time for analysts to find information and draw conclusions - Prototype operability + demonstrated scalability Data Fusion/Sensor Integration Software (THIS SECTION IS A WORK IN PROGRESS!) - Build solution that integrates with current systems (e.g. GCCS) - Work with PMs and key influencers to determine optimal funding/dissemination avenues - Deploy prototype, confirm buy-in and update features - Scale deployment, improve product as necessary Fixed - Buying proprietary data - Software tools - Hardware evaluation + prototyping equipment - Evaluation of commercial products Prototyping - Existing sensor platforms and feeds - Existing deployment platforms - Academic research - Existing data fusion platforms Scaling - Available commercial + military data - Existing database tools (Palantir, AWS) - Need commanding officer to confirm decision-making benefits - Need intelligence officers from ONI / N2 and operators from N3 to confirm effectiveness of insights - Need IT approvals to integrate into systems - Need support of commercial partners if want to leverage their platforms Beneficiaries Mission Achievement Mission Budget/Costs Buy-In Deployment Value Proposition Key Activities Key Resources Key Partners Military - 7th Fleet + designated sponsor - Naval Postgraduate School (NPS) - Office of Naval Research (ONR) - Acquisition Personnel Commercial - Distributed sensor platform companies (i.e. Saildrone, AMS) - Data analytics (i.e. Palantir, Google) Academic - Universities (i.e. University of Hawaii) - National Labs (Lincoln Labs, Sandia) Other - IUU fishing + anti-smuggling stakeholders (i.e. Coast Guard, PNA) - Disaster relief agencies Mission: Enabling Rapid, Well-Informed Decisions from Heterogeneous Data Testing - 7th Fleet assets for pilot - Research barge - Access to model analyst data interface - Access to sample incoming sensor feeds Variable - Travel for site visits, pilots - R&D personnel - Manufacturing/ Development IMPROVE TACTICAL AND STRATEGIC DECISION MAKING VIA BETTER DATA HANDLING (1) Rapid Strategic Decisionmaking via Improved Reporting (2) Improved Tactical Decision Making via Enhanced Information Sharing (3) More Effective Analysis via Searchable, Visualizable Data Integration (4) Predictive Intel and Alerts (e.g. Machine Learning) ENHANCE INCOMING DATA STREAMS (1) Improved Collection of Existing Data Streams (e.g. Fishing Broadcasts) (2) Painless Incorporation of Multiple New Sensing Modalities

8. Products & Services Timely data Good UI/UX for presenting data Streamlined reporting process Customer Jobs Gains Pains Gain Creators Pain Relievers Good UI/UX Platform incorporates more data streams Platform is robust and can handle drop out of data streams Allocate assets Identify, eliminate threats Predict hot spots Safety of team Projecting peace, stability in region More informed decisions Faster decisions Poor quality/lack of data Latency of data -> insight Admiral/Strategic Decision Maker Value Proposition Canvas

9. Products & Services Contextualized, object-oriented database Algorithms for processing, analyzing data Ability to search for trends across database Integration of disparate data sources Automation of data analysis Improved UX/UI Predictive notifications Customer Jobs Gains Pains Gain Creators Pain Relievers Contextualized, object-oriented database Compatible data format Incorporate multiple data streams Collect & analyze data Communicate findings Piece together contextualized awareness More actionable insights Faster identification & response times Easy-to-use Incorporation of context is manual/mental Poor quality / lack of data Latency of data -> insight Analyst (N2) Value Proposition Canvas

10. Products & Services N/A Actually a common operating picture! Customer Jobs Gains Pains Gain Creators Pain Relievers No hardware to deploy so no risk of asset or personnel loss Fewer change orders Utilize assets and human capital in order to obtain ISR data on adversary or regions of interest Better allocation and deployment of assets High manpower, time Operator error Safety concern for deploying in unfriendly territory Struggle to redeploy systems on short notice (<12 hours) = frustration Operations (N3) Value Proposition Canvas

11. MVP

12. MVP

13. MVP

14. Questions?

15. Customer Workflow

16. Customer Workflow N2 N3 N2 (“owns” the intel) N3 (“owns” the assets) Ready-To-Use Data Deployment Data Acquisition Data Analysis Data Order/Decision

17. MVP (1 week ago) AIS Weather

18. MVP (1 week ago) AIS Weather

19. MVP (1 week ago) AIS Weather

20. Data Acquisition Contextualized Database MVP (2 weeks ago) Deployment Last Month Today Object-oriented Database Query - What data is most useful to capture? - What sensor modalities can capture? - What products exist? - What deployment options exist? - What is easiest to deploy? - What is “good-enough” time to data acquisition? - What is the deployment process? - Is .kmz format all that is necessary for compatibility? - What do companies like Palantir do today?