tag:lazyspark.com,2013:/posts lazyspark 2016-11-02T12:58:19Z Sawan Ruparel tag:lazyspark.com,2013:Post/1104878 2016-11-02T12:57:45Z 2016-11-02T12:58:19Z Learning Impedance mismatch

The different forms of learning new content would be reading, listening, watching video or practical. What is the end result of the form of learning with the output - pretty much same. You learn how to perform a specific activity. An argument around reading vs practical is that practical will get you more proficiency at performing the activity. But the objective of learning is the same. 

Just imagine the scene from Matrix where Trinity was learning to fly.It was just in few seconds that she got the download. I believe that there must a certain way to accelerate the learning time. 

Tim Ferris in his four hour chef book and other preaching mentioned about meta learning. Learning how to learn. But again the toolset available to learn something is pretty old. We are limited by our sensory experiences to learn something. Can we just not take a pill or do some external process to make the neuron connections to learn something new? 

Going back to the title of the blog I believe that there is mismatch between information created and information that can be consumed. It is just too overwhelming. We cannot consume information fast enough and there is this impedance mismatch. 

Sawan Ruparel
tag:lazyspark.com,2013:Post/1099893 2016-10-20T12:16:50Z 2016-10-20T12:16:50Z Virtual assistants - Amazon Alexa, Google Home and Jibo

Let's look at virtual assistants available out there. What is available? and What are the ways to extend them? and Is there a way to compare them? 

Available virtual assistants 

Google Home

Google has all my data. Google Now just works. It would be like an extension to my current adopted Google ecosystem (voluntary or involuntary). So while in home if Google has more data to help me answer some questions like "Has Molly responded to my email on request for recipe?" would be great. Though above question is quiet complex Google has all necessary data to answer the question like all mails related to/from molly, mails in context of request for recipe and any recent reply from Molly. Also with Google Timeline I think Google is like a my personal robot following me everywhere helping me in my tasks and it is just getting better. Ofcourse I will be expecting more than what is possible but in few years it should be able to meet expectations. As of now it has standard integrations and the way features are divided in Google Home is through different categories like Enjoy Entertainment (TV and Audio Streaming Control), Get Answers, Manage tasks, Plan your day (calendar integration), Control home and others. 

Amazon Alexa 

Alexa is coupled with Amazon Ecosystem and works best with Prime subscription. Amazon Alexa works great (I have one and use it all the time for basic tasks at home) and only product out there is market available right now to buy and use. It has integration with lot of new devices. The problem is it doesn't have lot of data about me, it is not intelligent enough to suggest things like Google Now cards and the conversations with Alexa are very inhuman.  

Jibo (haven't heard of it? - it is very interesting one)

Jibo if you haven't heard of it is outcome of research at MIT Media Lab. It looks good has camera (not sure if anyone is up for having robotic cameras in house) and decent screen (I think it is projection screen). From the intro video I couldn't figure a lot but the company has been around for long enough to come up with right hardware for the job. It seems to be able to track people and follow people (through camera) thus can have interesting applications like while you are having a skype call the camera can follow your face. It can detect people in house which Alexa fails to address. There is an elaborate Developer API which we will explore in following posts. Overall Jibo merged with Google Home can do a lot. I wish Jibo gets acquired by Google and they make a hybrid device.

Unlike apps for mobile phones the things for virtual assistants can be called as skills. Is the platform similar - for this we would probably have additional posts.

Sawan Ruparel
tag:lazyspark.com,2013:Post/895678 2016-10-17T02:27:10Z 2016-10-17T02:27:10Z Lingo Inline

Learn new language while you are reading emails, reading news on the web. I had an idea that I had worked on quarters back called Lingo Inline. Developed an extension for outlook that will read email and modify few words in new language you are trying to learn. 

Why email? Because I am on email always and hooked to it.

Extension to outlook? Because outlook allows building extensions and I am on outlook for my emails. 

How it works? It used the outlook API. Allows read access to the email with the extension. I pick up few words from the email and show in the extension tab. 

Video from when it was developed -

Sawan Ruparel
tag:lazyspark.com,2013:Post/884174 2015-10-27T02:52:02Z 2015-10-27T02:52:03Z Use API or build the functionality
Should I use an API or build the functionality? Depends :) an architect would say.

Questions to ask before you choose
Is the API functionality core to the business function?
Do you need the API to be highly available?
Are there any compliance issues protecting the data flow in your application?
Is the API functionality available from multiple vendors?
In house expertise available to create the API? (NLP API)
Time to market?

Also a good practice should be to wrap the API into a vendor independent interface before using it in your application. So that you have flexibility to swap vendors or implement your own if you need to.

Sawan Ruparel
tag:lazyspark.com,2013:Post/892357 2015-10-27T01:03:15Z 2015-10-27T01:03:15Z Design of domain specific database technologies

Is it required to design a database technology specific to a domain? A typical example of this maybe design a database which has Healthcare Ontologies as a supported datatypes. Common database operations or database functions can be healthcare specific for coding, grouping and billing.

Today healthcare standards are closed and ruled by organizations like Optum, 3M and AMA. They charge for the use of code libraries, standards and standard groupers. I think there is an opportunity for startups to understand and disrupt this space.

Not sure if it applies to other domains like Retail, Banking, Education, Gaming, Automotive, Operating System Vendors and Telecom vendors. As we have been successfully able to map the data into pure relational form. Like Sales, Product, Geographies, etc. but then there is connections data which can be expressed in the form of graph and so we have graph db.

For scientific data we have something called SciDB. But again the database is designed by the type of functions and not the domain. The databases are more generic and try to offer generic type of language to perform data operations and not a DSL. In a healthcare database a specific operation or query may say - Group Claim(Id) for DRG

If someone reading this blog comes across a domain specific database please post a link

Sawan Ruparel
tag:lazyspark.com,2013:Post/664312 2014-03-16T02:57:33Z 2015-02-02T14:28:57Z Ideas from Telehealth and Hospital CIO track @ MIT Med GrandHackFest


Data points on Telehealth - 
Mass General admission data - 1.5 million outpatient visits, 1.5k beds, 47k inpatient admissions. Potential of 30% as televisits
Video conferencing - trust is important - doc presence will help get better

Current departments - 
Telestroke - first found in 1996
Teleneurology - virtual visits 
Telecardiology - virtual visits 
Telepsychiatry - virtual visits 
Teleburns - virtual visits 

Barriers for adoption -
1. EHR implementation 
2. Providers are worried 
3. Regulations 
4. Security
5. Reporting 
6. Don't get paid to see people remotely 

Competitions for providers -
1. Minute clinic
2. Over the counter med
3. Community 

Ideas - 
Build a platform for monitoring - visualization tool
Decrease the efforts to adopt televisits
Increase integration and insight at the point of care


Pain points -
1. Meaning full use issues - good in 2015 
2. Break monolithic EHR 
3. Usability in EHR 
4. Patient and family engagement tools are poor / value proposition 
5. Re-represent data  
6. ICD 10 BILLING ISSUE - put in wrong codes 
7. Privacy issues - HIPAA  

Ideas - 
Doctors are overwhelmed with data - Beth Hospital has 3 PetaBytes of data 
Record of the future - Wikipedia for patient

Sawan Ruparel
tag:lazyspark.com,2013:Post/652641 2014-02-10T02:48:02Z 2014-02-10T02:48:03Z BI tools transformation from passive to active analysis

Microstrategy, Tableau, Spotfire, Qlik and the list goes on. BI tools offer software to understand and report data. But can you really understand data without under standing the business well, so the tools don't provide anything to do with business. In college I learned Matlab. Matlab had the set of tool boxes and it allowed to understand the subject better like DSP or others. 

So that's the difference between active and passive analysis tools. Passive analysis tools like the current popular once leaves it upto the user of how they want to use it, whereas active tools can provide you toolbox. 

The BI tools needs to have plugin ecosystem to create domain specific BI tools. Like Healthcare BI tools, why is the BI market hesitant to create that? Or why is the BI market hesitant to open up there products as platforms. From my experience I see that the BI tools with such an ecosystem will be successful especially with small-mid companies who cannot afford to have their own experts in the domain to provide classification and other basic data analysis. 

But the tools who have such an ecosystem like SAS are too expensive for small enterprises. Probably then the solution is Web BI tools. Lot of them have emerged in last few years definitely few of them might have thought to be platform for the specific domain instead of being generic BI tool. 

 Waiting for more active analysis BI tools. 

Sawan Ruparel