Author: Ashish Kakran

Here is why enterprise data leaders care about the Modern Data Stack

A version of this blog appeared in TechCrunch: Is modern data stack just new wine in an old bottle?

Remember the cable, phone and internet combo offers that we used to receive in our mailbox? These offers from cable companies are highly optimized for conversion. The type of offer and the monthly price can vary significantly between two houses right next to each other or even between different condos in the same building. I know because I used to be a data engineer once and built Extract-Transform-Load (ETL) data pipelines for this type of offer optimization. Part of my job involved unpacking encrypted data feeds, cleaning them to remove rows or columns that had missing data and map the fields our internal data models. The clean, updated data was then used by our statistics team for modeling the best offer for each household. This was almost a decade ago. Now take this process that I described, run it on steroids for 100x larger datasets and that’s the scale that mid-sized and large organizations are dealing with today.

Take for example, a single video conferencing call can generate logs that require 100s of storage tables. Cloud has fundamentally changed the way business is done because of its unlimited storage and scalable compute resources at an affordable price. A simple comparison between the old and modern stack looks like this:

Why do data leaders today care about the modern data stack?

  • Self-service analytics: the citizen-developers want access to critical business dashboard in real-time. Their desire is (Read more...)

Thomvest invests in Isovalent: transforming cloud networking, security and observability

Thomvest Ventures is excited and thankful to have led the Series B round for Isovalent, the company behind the open-source Cilium and eBPF projects. Isovalent is disrupting networking, security and observability for large-scale cloud native deployments — from 100s to quite literally millions of containers in the real world. Developers and platform teams love Cilium which is now seeing rapid adoption within the open-source community — over 450 contributors, over 12,900 Github stars and an active Slack community of over 12,500 engaged users

When we first spoke with Dan Wendlandt, we were immediately impressed with the vision of the company he is building. Having been a Product Manager for a distributed-transaction-tracing product, I knew how hard it is to instrument an application and then collect data without adding performance overhead. I was fascinated with the power of eBPF to securely collect metrics data from the linux kernel without making any application changes. After a follow-up technical deep dive with Thomas Graf we were blown away and felt lucky to have met the team.

Web applications today don’t just support the business, they are the business! If the application is down or slow, the revenue impact in the SaaS subscription world is immediate. Technology architectures have evolved to enable quick software deployment, fast bug fixes and rapid iterations for best end-user experience. Modern apps are built using microservices, are hosted on a multi/hybrid cloud and heavily utilize containers and Kubernetes orchestration platforms.

This simplicity in architectures can however cause management complexity at scale when it comes to (Read more...)

Thomvest invests in Opaque Systems — from friendship to company-building

Thomvest invests in Opaque Systems — from friendship to company-building

Thomvest is thrilled to announce our Series A investment in Opaque Systems, the company behind open source MC2, a platform for collaborative analytics and artificial intelligence ( AI) at scale. Confidential computing is a market that is ready for takeoff — Gartner estimates that by 2025, 60% of large organizations will adopt privacy-enhancing computation (PEC) for processing data in untrusted environments and (multiparty) data analytics use cases. Opaque Systems was founded by an incredible team — Rishabh Poddar, Raluca Ada Popa and Ion Stoica — who have the right skills to solve this incredibly technical challenge.

Our journey with Rishabh, Opaque’s co-founder and CEO started many years ago during his PhD days at UC Berkeley. My interest in one of his published papers led to a memorable early-morning coffee meeting, when he was a member of Berkeley RISELab whose mission is to develop technologies that enable applications to interact intelligently and securely with their environment in real time. Rishabh’s excitement about the confidential computing space is infectious and we knew right away that we had to be a part of his journey whenever he decided to launch his dreamboat. We are thankful that time has arrived!

AI and machine learning (ML) techniques are increasingly being adopted by enterprises today to make business-critical decisions. For continued accuracy, machine learning needs a large amount of data which today is stuck in organizational silos. Sharing confidential data is a bottleneck for large enterprises in regulated industries who want to (Read more...)

Here’s where MLOps is accelerating enterprise AI adoption

A version of this blog appeared in TechCrunch: Here’s where MLOps is accelerating enterprise AI adoption

Why is MLOps important?

In early 2000s most business-critical software was hosted in privately-run data centers. Initially skeptical, enterprises moved their critical applications to the cloud. DevOps fueled this cloud adoption as it gave decision makers a sense of control over business-critical applications hosted outside their own data centers. Today, enterprises are in a similar phase of trial and acceptance of ML in their production environments and MLOps is acting as an accelerator. Similar to cloud-native startups, many startups today are ML-native and offer differentiated products to their customers. But a vast majority of large and mid-size enterprises are either in trial phase or just struggling to productionize functioning models. Here are some key challenges that MLOps can help with:

Cross-team ML collaboration is a must but tough to accomplish

A model may be as simple as a churn-prediction model, or as complex as the one determining Uber or Lyft pricing between San Jose and San Francisco. It is an incredibly complex task to create a model and enable teams to benefit from it. In addition to a large amount of labeled historic data to train the model, multiple teams need to coordinate to continuously monitor the models for performance degradation. There are three different core roles that are involved in ML modeling but each one has different motivations and incentives:

1. Data Engineers: these are trained engineers who excel at grabbing data from multiple (Read more...)

Thomvest Ventures Invests in CyCognito, the External Attack Surface Management Leader!

We are thrilled to announce our Series C investment in CyCognito, the company helping the world’s largest enterprises quickly prioritize, investigate, and respond to potential security risks. The company brings to life what Gartner calls External Attack Surface Management: making visible what attackers see when looking at organizations from outside-in. Our first meeting with Rob Gurzeev, the founder of CyCognito, felt like magic. Rob previously led teams at Unit 8200 of the Israeli Intelligence Corps. In venture capital, you only get to have a handful of such meetings each year with a visionary founder who has the vision to see a significant market opportunity, the technical depth to build a great product, and the leadership experience to build a great business. Rob once ran a team that used to think like nation state attackers and developed defenses against them, giving him unique experience that stood out to us. We are excited about the mission that Rob is on!

Gartner estimates that 50% of the breaches can be attributed to unknown and unmanaged assets. Such assets that make up the external “attack surface” could be servers, API endpoints, code repositories, Kubernetes clusters, containers, virtual machines, Amazon S3 buckets, open-source components and much more. But this previously unknown attack surface must be contextualized with all known assets to get a full, prioritized and actionable view of your entire surface. CyCognito can do this by extensively using Artificial Intelligence (AI) — specifically Machine Learning (ML) and Natural Language Processing (NLP) — to automatically discover and security test the (Read more...)

Trends impacting DevOps in 2021

Engineers at fast-paced startups often wear both dev and DevOps hats. Early in my career, as a developer at a sentiment-analysis startup, I was tasked with writing scripts to spin up servers, configure them, install necessary software and deploy the code we wrote. My custom shell scripts, python scripts and cron jobs did all the heavy lifting so that we could reliably deploy software and scale resources on AWS. Fast forward to today, that problem of confidently building and securely deploying software has become an enormous challenge at scale.

The cloud native application stack

A simple looking ecommerce app today can easily be a big set of coordinating microservices that are powered by containers which are orchestrated using Kubernetes hosted on multi/hybrid cloud. This combination enables faster delivery of code, cheaper hosting bills and more efficient utilization of server resources.

While the application components certainly became smaller over time, organizations today are tasked with managing tens of thousands of containers in a production environment. In a 2020 CNCF survey, 92% of respondents said they use containers in production, a 300% increase from just 23% in the first survey from March 2016. Such coordination complexity at scale is being managed by software delivery pipelines. Just like Ford revolutionized automobile industry with its assembly line, these integrated DevOps products are massively impacting verticals like finance and insurance.

DevOps trends

Developers are now decision makers and influencers

We are excited to see the impact of developers in the selection of tools. The top-down (Read more...)

Bolster and Shiftleft recognized as 2020 SINET 16 winners

Each year SINET evaluates hundreds of innovative cybersecurity companies from around the world to select 16 of the most innovative companies. We are delighted that our portfolio companies Bolster and Shiftleft are amongst the 2020 SINET winners! This is a great recognition of the team and the differentiated platforms of these cutting edge-technology innovators as they address complex cybersecurity pain-points for their customers .

A massive problem faced by global enterprises is online fraud. The Bolster platform detects phishing and fraudulent sites with high fidelity in real-time and automatically takes them down. Organizations like Zoom, Uber and many others use Bolster to protect enterprises from bad actors on the internet and protect their valuable brands.

Another challenge faced by enterprises is ensuring code-related security as Agile teams push software releases to production many times a day. Shiftleft built a nextgen platform which analyzes code 40 times faster and 3 times more accurately, with minimum false positives compared to legacy static analysis tools. Trusted by Intuit, Cisco etc, Shiftleft’s extreme speed and accuracy enables developers to find vulnerabilities before releasing their software .

We are excited to be part of the journey with Bolster and Shiftleft teams as they build very successful companies. Congratulations!

Bolster and Shiftleft recognized as 2020 SINET 16 winners was originally published in Thomvest Ventures on Medium, where people are continuing the conversation by highlighting and responding to this story.