Welcome Sreeni Iyer! SlamData’s New SVP of Engineering

Chris Dima

Chris Dima Marketing
Mar 26, 2017

Sreeni Iyer: "SlamData is novel, it works and most importantly it brings immediate value to companies who are amassing this type of data on a daily basis and gives you this entirely new way to solve the problem."

Why Did You Decide To Join SlamData?

Welcome to SlamData Sreeni! Let’s get right into it. First question. Standard, of course. What was it about SlamData that interested you the most? Why did you decide to join the team?

Well, the first thing to draw attention to is that the morphology of underlying data is mutating — structured data volumes are exploding, but semi-structured and unstructured data are exploding exponentially more. The industry as a whole has not come up with an adequate way to access and therefore maximize value from this type of data. Yes, via the NoSQL, Hadoop ecosystem and related innovations, we have a handle on scale issues associated with data capture, ingestion and persistence at a reasonable cost, but two key questions along with a few others aren’t yet met with good answers. One is “Let me ask questions directed at these massive, heterogeneous, distributed, low- to high-entropy datasets and get a human-consumable answer in reasonably quick time.” Two is “I do not know what questions to ask of this data, so compute and highlight correlations, predictions and ‘latent insights’ and then let me explore them.

Sreeni Iyer, SalmData's new SVP of Engineering, speaking at the 2015 IPS conference in Montreal.

Sreeni Iyer, SlamData’s new SVP of Engineering, speaking at the 2015 Neural Information Processing Systems conference in Montreal.

Get Immediate Insight Into NoSQL!

Start Your Trial

SlamData is at the forefront of solving these issues properly and not in a “hand-wavy checkbox feature” manner. As an example, with the first question, today one typically ETLs or ELTs data from disparate stores (so-called ‘polyglot persistence’) into a canonical structured shape located in a central store before analytics can happen and answers returned to the end-user.

This data transformation and re-shaping takes engineering time and effort and makes the data-to-insight journey less agile, treats these source/stores as “dumb ingest end-points” (and hence minimizes ROI on those investments) and requires “copying data” as the only way to solve the problem. Whatever happened to move compute to the data and not vice-versa?

Further, Catch-22’s abound within this methodology since end-users are unaware of what data exists. They are often unable to define what is to be “extracted”. The second question needs data scientists, data engineers and analysts to collaborate. This equals time and money, and in some cases data perishability imposes tighter constraints around how long one can wait for a large cast and crew and tool sets to line up together.

I thoroughly investigated this problem in the context of a vertical in my previous startup, Quad Analytix, which focused on getting value out of unstructured/semi-structured data and generating insights for e-commerce merchants and marketers. More recently, as Entrepreneur-In-Residence at Shasta Ventures, I evaluated quite a few analytics companies. What I found consistently was that they all required the aforementioned re-shaping of diverse data sources/formats into a tabular form and loaded in a central store before slapping on analytics and visualization tiers.  The innovation was not at the core but at the edges — i.e. making the ETL “suck less” or improving the ingest/analytics capability in some ways or making the visualization slicker.

And for companies with disparate data, SlamData’s SQL2 becomes lingua franca and its compiler-based, push-down approach, makes data easy to access without the need for honking, large, central, ingest infrastructure.

So that was the attraction to SlamData — it innovated at the core and solved this data-to-insight “agility” problem. It solved it with new algebra along with an efficient implementation on top of that, to deliver its compelling, compiler-based approach. End-users express declaratively the data they need via SQL2 (which is SQL with some additional expressiveness and which the entire industry is familiar with). What results truly appeals as elegant to an engineer — moving compute to the data and doing “push-down” semantics versus always needing to “pull-up” and reshape via ETL or ELT.

Altogether, SlamData is novel, it works and most importantly it brings immediate value to companies who are amassing this type of data on a daily basis and provides an entirely new way to solve the problem. Now, when battling beasts in the data-jungle, who doesn’t want another set of special arrows in one’s quiver?

The ROI on SlamData?

Thanks. That sums it up! Now think ahead two years to our clients who are using SlamData. What is their life like? What is the return on investment for adopting SlamData versus a legacy or home-grown solution?

Let’s try to answer this with an analogy. Let us suppose I provide you compelling content via many massive libraries, but to access this content you need to speak/read Dothraki, Na’avi, Klingon, Valyrian and Vulcan.

Maybe the creators had fun creating these languages and a linguist or language aficionado may like this state of things, but what about the rest who are comfortable with English and care only about the “compelling content”? In this scenario, we could adopt a “do it yourself” approach and hire an army of linguists who learn these languages and answer questions we want to pose against this content (and often the language barrier means we are unsure of what exists in the first place, let alone knowing how to formulate pointed questions to pose). Or, one can physically translate these books into English and stack them in another library — then and only then can we ask questions off this content.

Both approaches sound expensive, seem far from agile and are unwieldy. What if,  however, you had a babelfish translator that automatically transcribed English into Dothraki, Klingon and Vulcan — that would be magical! You could start answering “ad-hoc” questions on your own, without needing this army of linguists (datastore API experts) or an English library (tabular store in a central location after mapping/ETL/ELT/data processing & prep.) in between you and the original content.

Slamdata is the Babelfish of modern analytics in this world of polyglot persistence (RDBMS, NoSQL, Hadoop, Apps that spit XML/JSON) translating from SQL2 to plain-old SQL (RDBMS), MongoDB-API, XQuery (Marklogic), N1QL (Couchbase),  Map-Reduce/Spark (HDFS). It pushes-down compute to these stores, so you don’t need the massive tabular store in a central location and you maximize your investments in these stores’ computate engines.

All this is done via Quasar — an open source library maintained by SlamData that exposes this power via its interface and making the answers “human consumable”. So in a way, this is the state of the mainstream today, there is compelling data and a map to a treasure in this multi-lingual library somewhere, but to get to it, either you are well-versed in MongoDB-API, N1QL, XQuery, Map-Reduce/Pig/Hive/Tez/Impala and a dozen more or you invest in today’s analytic solutions, which build you a library with content in “English” (Ingest/Map/ETL/ELT + Shape/Re-shape/Prep approach).

And for companies with disparate data, SlamData’s SQL2 becomes lingua franca and its compiler-based, push-down approach, makes data easy to access without the need for honking, large, central, ingest infrastructure.

Where SlamData provides a tremendous amount of value is its ability to provide agile analytics (sometimes called ad-hoc or flash analytics). SlamData is performant with inbuilt, specialized optimizers. And for companies with disparate data, SlamData’s SQL2 becomes lingua franca and its compiler-based, push-down approach, makes data easy to access without the need for honking, large, central, ingest infrastructure.

Companies running SlamData would have a transformed analytics practice — deriving quasi real-time insights from their most complex data. And doing it in a lightweight way that doesn’t consume time, money, serious infrastructure or require training. That provides a huge benefit in terms of simplification and material competitive edge via reduced cost and faster time to market.

Get a Jump Start!

Work with us and we'll jump right over the hang ups you've been facing and go right to Insights.

Immediate answers. Check!
Flexible reporting. Check!

No Mapping! No Coding!

Learn More

How Will the Market Respond?

Moving data is deeply ingrained in the industry. It’s expected. Mapping data is the same way. People never expected there to be a way around these things. How quickly can things change?

Yes the old ways are ingrained and change often appears threatening initially unless it truly presents a new and better way. SlamData’s compiler-based approach with push-down semantics is based on the newer mathematics of MRA (multi-dimensional relational algebra). It’s market-changing — SlamData definitely delivers a revolutionary approach. It delivers with semi-structured JSON/XML of arbitrary length & nesting complexity; and shines where naive flattening approaches fall flat – pardon the pun 🙂 The naive approach is what many of today’s players such as a Tableau, Qlik, Dundas BI and others of that ilk adopt via 5-year-old JSON/XML connectors.

There is, however, no imposing a new world order on an organization — it’s not “my way or the highway”. It’s easy to implement SlamData on new or existing projects. It’s open source, backed with enterprise-relevant offerings, browser/tablet-based and doesn’t require engineering involvement. If you still feel compelled to go down the traditional ETL route for more well-defined problems, then do so — this is an additive, complementary approach, which gives you additional super-powers especially with ad-hoc analytics vis-a-vis your semi/unstructured data. So, really there is no barrier to adoption — try it and experience its power alongside your more traditional approaches and see for yourself, why we are so excited and trying to share our enthusiasm about this.

It’s almost a conundrum, because it is revolutionary, but it relies on SQL, the 30-year analytics powerhouse.

We’re about removing the “No” from NoSQL, because we believe “NoSQL” is a misnomer. The data revolution meant  to say “not a Relational DBMS-based approach”. While “NoSQL”  solved some scalability and schema-rigidity issues, it threw the baby out with the bathwater. SQL (Structured Query Language) by itself is a good idea — at the high-level it says “keep the access declarative irrespective of how/where the data is actually stored”. The industry, it seems, forgot that and introduced dozens of custom APIs and left end-users with compromised accessibility.

The only way to access the data is usually via custom APIs and approximations. We get that for semi-structured JSON, for example. In addition to traditional primitives, we now have lists and maps involved or other complex combinations — but subsume this burden of navigating this complexity instead of passing it  to the end-user which is the raison d’etre for SlamData.

This is counterintuitive to what’s happening in the real world where we are increasingly data-driven even in our personal lives and of course in our professional lives, where businesses want to supplant intuition with data-driven insights. Companies want more ‘data democracy’… hence the birth of ‘citizen data scientists’.  As an end-user, one may not have all the technical skills (maybe SQL is the only thing I am well-versed at or maybe not even that), but I still want access to the data and ask it questions,  so I can back my decision with insights from the data and reduce risk to the business.

It’s really about enabling those citizen data scientists, and not denying them the access to the data or making them jump through hoops,  because of the way the technology was implemented. SlamData is a single way to access your data for analytic payloads — it’s making NoSQL as easy as RDBMS. SlamData is the next-generation analytics engine for data with “higher entropy” than you find in structured forms. Sometimes an “Undo” is what is needed to make progress on a journey. So, no(noSQL) = SQL2 is “Back to the Future”!

The fastest, most agile way to
do analytics on disparate data.

Removing the Number of Hoops

So if you’re a CTO and you’re running a number of data sources, and then your developers are saying add X new database, then you’re saying “Now I need another reporting solution or analyst…” So if SlamData simplifies this, then the value of having access to data, and having the ability to easily add new data sources, goes through the roof?

Yes, exactly. It’s all about removing the number of hoops that one has to jump through in order to get the value from the data, without constraining what stores a given application will persist to. Application engineers should be free to choose their online persistence store, based on the demands of their application. Analytics should become an orthogonal concern, without all this complex hoop jumping.

Today, the hoops include ETL/ELT and mapping; data preparation/processing  you have to execute to transform/re-shape the data from one form to another; Learning/getting used to store specific APIs and multiple-tool sets; Figuring out the ML/NLP approaches to adopt to squeeze out more; Stitching in the visualization tier (charts) and the form elements on top of that, so that the user can interact with a browser-based or mobile application and get a question answered.

SlamData is about simplifying all of that and allows an end-user with SQL knowledge or even with none to build a data-application in minimal time and insulate themselves from data-entropy and how/where/which store the data is persisted in. Its revolutionary technology that eliminates the need for an army of engineers and 3-6 months to start answering questions.

The Focus For the Next Six Months

What is your focus at SlamData over the next six months?

The focus over the next six months is to basically harden the core execution engine and add capabilities that far exceed the industry’s capabilities today. Imagine being able to join a MongoDB collection to a Marklogic table or a Twitter feed on the fly and query it via SQL! Then polish the UI/UX and add more “connectors” i.e. translators to a given store’s dialect.  So, it’s to make sure we are production-ready and focus on performance,  scale in a variety of usage scenarios.

Rugged would mean being able to go from download to value in minimal time for not just a department app, but for mass adoption across the entire enterprise.  We’re also putting efforts in terms of polishing the UX and the UI — so far our focus has been on solving the complex mathematical problems at the back end, and exposing that power to the end-user, but we haven’t focused that much on the visual elements — so  we are rectifying that. We’ll also be adding a lot more connectors. As you know, we have a connector for MongoDB that’s been tested in production under various loads, and it’s doing well. We have connectors now for Couchbase and MarkLogic and HDFS (via Spark) and more are on the way.

The Team

And what about the team to support this?

We doubled our back-end team and our front-end team this first quarter. What we have here, without sounding boastful, is easily one of the best “functional programming” teams on the planet — Scala, Purescript, and Haskell in play — very committed to solving problems with minimal “debt”. This has helped attract fantastic talent and encouraged OSS contributions — we are firmly committed to the open source model). In addition, we’re also hiring folks who can focus on performance testing and make sure that we are able to repeatedly prove for ourselves, that SlamData easily outshines other approaches in the market, today.

Thanks Sreeni! We’ll check in with you in two months and get an update from you.

Thanks Chris. Looking forward to sharing what we accomplish in the short term.

News, Analysis and Blogs

Send Us A Message

Get In Touch

(720) 588-9810
[email protected]
1215 Spruce Street, Suite 200 Boulder, CO 80302

Connect With Us

© 2017 SlamData, Inc.
Stop Wasting Time With Data Prep and ETL!
Start Your Trial