Algorithmia founder on MLOps’ promise and pitfalls

All the sessions from Transform 2021 are available on-demand now. Watch now.

MLOps, a compound of machine learning and information technology operations, sits at the intersection of developer operations (DevOps), data engineering, and machine learning. The goal of MLOps is to get machine learning algorithms into production.

While similar to DevOps, MLOps relies on different roles and skill sets: data scientists who specialize in algorithms, mathematics, simulations, and developer tools, and operations administrators who focus on upgrades, production deployments, resource and data management, and security. While there is significant business value to MLOps, implementation can be difficult in the absence of a robust data strategy. Kenny Daniel, founder and CTO of Algorithmia, the company behind the enterprise MLOps platform, spoke with VentureBeat about the buzz around MLOps, its benefits, and its challenges.

This interview has been edited for clarity and brevity.

VentureBeat: How does MLOps work?

Kenny Daniel: MLOps is applying the lessons of DevOps and software engineering best practices to the world of machine learning. MLOps includes all the capabilities that data science, product teams, and IT operations need to deploy, manage, govern, and secure machine learning and other probabilistic models in production. MLOps combines the practice of AI/ML with the principles of DevOps to define an ML lifecycle that exists alongside the software development lifecycle (SDLC) for a more efficient workflow and more effective results. Its purpose is to support the continuous integration, development, and delivery of AI/ML models into production at scale.

We break down MLOps specifically into 10 core capabilities across the Deployment and Operations stages of the three-step ML lifecycle (Development, Deployment, Operations). Across the Deployment phase of the ML lifecycle we have:

  1. Training integration — broad language and framework support for any DS tooling.
  2. Data services — native data connectors for popular platforms, as well as permissions and access controls.
  3. Model registration integrated with your docs, IDEs,  and SCMs, with searchability and tagging so you know the provenance of all your models in production.
  4. Algorithm serving and pipelining — allowing for complex assemblies of models required to support the app — this should be hands-off maintenance.
  5. Model management — how you control access for version management, A/B testing, source and licensing control, and build history management.

Across the Operational phase, there are also five core capabilities:

  1. Model operations — which is how you control usage and performance in production, includes approval process and permission control.
  2. Infrastructure management, which includes fully automated infrastructure, redundancy, autoscaling, on-premise, cloud, and multi-region support.
  3. Monitoring and reporting — visibility into the “who, what, where, why, and when” of MLOps.
  4. Governance, logging, reporting, customer metrics for internal and external compliance.
  5. Security, across all stages, including data encryption, network security, SSO and proxy compliance, permission, and controls.

VentureBeat: The nature of the AI deployment depends on the organization’s maturity. In this case, what needs to be in place for an organization to be ready for MLOps? 

Daniel: MLOps becomes relevant when trying to get machine learning models into production. This will typically happen only after a data science program is established and projects are well underway. But waiting until the model is built is too late and will result in delays in getting to production if the MLOps story is not solved.

VentureBeat: What are common mistakes with MLOps?

Daniel: Leaving the responsibility on the individual data scientists to navigate the IT/DevOps/security departments on their own. This sets up a recipe for failure, where success depends on a specialized team navigating a completely different software engineering domain. We’ve seen a lot of companies that will hire teams of data scientists and machine learning engineers and set them loose building models. At the point where they’ve built a model and need to get it deployed and ready to handle production traffic, there are a number of things that need to be in place. These are things that are considered mandatory in the modern IT environment, not just for machine learning: source code management, testing, continuous integration and delivery, monitoring, alerting, and management of the software development lifecycle. Being able to effectively manage many services, and many versions of those services, is especially critical in machine learning, where models may be retrained and updated on a constant basis. That’s why it’s critical for companies to answer the question of “What is our MLOps story?” and what is the organization’s process for going from data, to modeling to production.

VentureBeat: What is the most common use case with MLOps? 

Daniel: Large enterprises use us for mission-critical applications. The most common use cases we see are those that are critical to scaling complex applications to gain agility, accuracy, or speed to market; anyplace where a faster transaction has a material impact to value. Merck, for example, speeds up the analysis of complex compounds for drug discovery and vaccine development. EY accelerates fraud detection by updating models more frequently and reducing false positives by over 30% with those better-performing models. Raytheon will support development of the U.S. Army’s Tactical Intelligence Targeting Access Node program.

VentureBeat: How has the advent of low-code/no-code helped/hindered MLOps?

Daniel: I am generally skeptical of low/no code solutions. The good thing is that because they are typically opinionated about the applications they produce, they often come with a solid MLOps story out of the box. The downside is that while they might be quick to get working on a simple demo, most real-world applications will have complexity that goes beyond what no-code tools can support. The customization becomes critical for applications in production.

VentureBeat: DevOps quickly went into DevSecOps as developers realized that we should be integrating security operations into development as well. Is there a security element for MLOps?

In our research, security, along with governance, is the top challenge that organizations face when deploying ML models to production. There absolutely is a security element for MLOps, and it is converging with more traditional data and network security. Enterprise-grade security is definitely something ML Engineers must consider as a first-order capability of any MLOps domain. I’m talking about data encryption at rest and in flight, unique model containment, API pairings, private and public certificate authority, proxy support, SSO integration, key management, and potentially air-gapped deployment support for high-security usage.


VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the subjects of interest to you
  • our newsletters
  • gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
  • networking features, and more

Become a member

Repost: Original Source and Author Link

Tech News

5 strategies every founder should use to defeat their detractors

This article was written by Joe Procopio and originally published on Built In.

Every startup is going to have its detractors. Every new idea, every revolutionary product, every unique strategy, and every best-made plan will find its natural enemy out in the wilds of the market.

When you’re an entrepreneur, you’re going to face armies of people who doubt you. At best, they’ll throw a cutting remark out on social media. At worst, they’ll do everything in their power  —  and they have a lot of power  —  to stop you from threatening their status quo.

Your detractor could be an incumbent company in a niche market that they’ve been taking advantage of for too long. It could be a potential customer’s “review” proxy, a tech or finance or other expert who is brought in for the sole purpose of poking holes in your solution.

The detractor I want to talk about is the “status quo hero”  —  an employee of a potential customer organization who has carved out an antiquated business process, made it intentionally complex and bureaucratic, and now manages an entire kingdom of inefficiency. These heroes usually lead teams of people carrying out the same costly tasks and making the same costly mistakes over and over again — because that’s how it’s always been done.

You have to kill these detractors with a combination of kindness and knowledge. You have to overwhelm them with shock and awe (kindness) and at the same time deplete their ammunition (knowledge).

This is never easy to do. Believe me, they’ve fought and won their battle many more times than you have.

But not more than me. I’ve been battling the status quo hero for a long time. Here’s what you need to know to get inside their head.

You are your detractor’s secret weapon

The first tactic a detractor will try is to get you off your game. Don’t get mad at them. Don’t complain about them. Don’t speak ill of them to anyone within their organization, even if one of their colleagues speaks up first.

I start here because I’ve seen this battle go the other way, where the founder or the sales lead was cocky, even confrontational, about the obvious necessity of their solution in the customer’s organization. You know what this does? It makes your customer think that you think that they’re stupid.

Try to close that sale.

No, after 20 years of fighting detractors, one thing I’ve learned is that logic doesn’t matter. Most people (probably you and me included) will do more work to keep things the way they are now, especially when they think their way is the right way.

If you’ve ever been in a situation where a friend of yours was struggling with a task, and you tried to help, and you maybe said something like “You’re doing it wrong,” or even “You’re trying too hard.” Well, you know how that goes over. Rarely well.

This is the bane of innovation. You can’t argue with a detractor, you have to empathize with them. You can’t tell them what they’re doing wrong, you have to get them to want to do it better. You have to win them over and become their secret weapon.

Understand your detractor by exploring the root cause

There’s one industry I’m very hesitant to do any work in, and that’s healthcare. I’ve done it before, and the industry is an entire detractor’s user group working in one building. I imagine even Band-Aids were vilified at some point.

But I ended up being an advisor for a startup in the healthcare space because the solution is so good it trumped all my negatives. This company constantly faces old-school healthcare people not wanting to use the solution and actively fighting against it.

To give you an idea of how deeply embedded the status quo is, just go to your doctor’s website right now and try to book an appointment. I bet 50 percent of you will just think,  “Yeah, I know what you mean.” And another 49 percent of you will think, “What website?”

How does a status quo like that happen? The same way it happens in every industry, and I’m here to tell you it’s technology’s fault. In healthcare, in particular, it’s decades of enterprise resource planning-style software replacing paperwork. Most of the time, that software ended up being more complex than the paperwork itself, and thus it made every process take longer, procedures became bloated, and huge swaths of friction got added to every single job.

That software was sold with the same promises that innovative startups are making today: “We’re from technology, and we’re here to help.”

Now, this is a niche example from a single industry, but the prognosis (pun intended) is the same for any product in any industry.

Here are five strategies I’ve developed to not only push back on detractors in a positive way, but also lay the groundwork to eventually win them over and become their secret weapon.

Credit: Built In

1. Stop selling technology

Technology can be a dirty word to some people, and again, that’s technology’s fault. You may be running the most technical company on the planet, but I assure you, you aren’t selling technology. So stop it.

What you’re selling is the benefit that the technology produces for the customer. It could be software that automates tasks, not digitizes them. It could be machine learning that eliminates the drudgery of simple decisions, not replaces human employees. Hell, it could be a carbon fiber reinforced polymer that makes a hammer easier to swing, not a damn carbon fiber reinforced polymer.

Your customers aren’t stupid. They’ll know that when you say “carbon fiber reinforced polymer” it’s because you think it’s scientific enough to go over their heads.

Technology is the primary weapon in your detractor’s fight against you. Sell benefits, not specs.

2. Get out of the training business

Any product that takes more than a couple hours to train for first use is probably bloated. No judgments here, I recently oversaw the delivery of a three-hour training session on some of our own software. In my defense, it wasn’t built for customer-facing access, because if it was, I’d have already had a walkthrough built.

Get into the walkthrough business. The less education you have to do before the customer gets results, the less ammo you’re giving the detractor in terms of the argument of removing complexity by adding more complexity.

It can also remove the complexity argument entirely. One of the products I launched with Spiffy a couple of years ago came with a three-inch by three-inch instruction sheet with four pictures and five words. It had about nine months of coding and technical development behind it.

3. Compartmentalize and conquer

The odds are that you can’t win over an entire organization at once, but if you find a champion who can translate the value of your product through the bloat of what the job has become, you have a decent shot. You’re basically fighting the detractors from the inside, which is insidious and awesome.

We did this at Automated Insights, a company that produced narrative reports like news articles from raw data. When we took our engine to the Associated Press to write their high-profile public company quarterly earnings recaps, we found an ally in a vice president in their editorial department. They championed our product’s value to journalists who thought we would be eliminating their jobs and at the same time didn’t believe our technology worked.

Yes, the double whammy. How could we replace their jobs if our tech didn’t work? Remember what I said about logic being a non-factor? This was that. When the correct information comes from the inside, it’s that much harder for detractors to stir doubt, because they’re calling out one of their own.

4. Incentivize change

A great way to beat detractors inside of your own organization is to incentivize change. This is not just handouts and kudos, it’s rethinking the way performance is measured.

We did this at Spiffy a few years ago when I launched an “instant upgrade” program to resolve problems we found on the spot. This program required our technicians to inspect each vehicle before they serviced it, which added about 10 minutes to each service. But simple math told us 10 minutes 10 times a day means one less service a tech could do in a day.

So we changed the way we incentivized their performance, tweaking quality above quantity, which led them to find an aggregated 10 other minutes during each service that weren’t necessary and eliminate that waste. Everybody won.

5. Take incumbent terms out of your lexicon

I’m advising another company that’s changing the way recruiting and hiring is performed. Their solution disrupts the process entirely, with quantifiable results. But it’s a total change in an area with a lot of detractors waiting in the wings to defend spending days on tasks that should take minutes.

To capitalize on the benefits of this new type of product and to disarm the detractors, I recommended the startup take out all references to incumbent industry terms and jargon. Those terms aren’t in the software, they aren’t in the marketing materials, they aren’t in the sales pitch.

This tactic, like all the others, resulted in an added bonus of making it clear that we weren’t the detractor’s enemy, we were their ally. We weren’t taking their jobs, or their power, or their inefficient kingdom. Instead, we were giving them better jobs, promotions, if you will, with more power in a more efficient and stronger kingdom.

Even the most hardline detractor will find it tough to argue with that.

Repost: Original Source and Author Link

Tech News

Google reportedly fired Margaret Mitchell, its Ethical AI Team founder

Google has evidently fired the founder and co-lead of its Ethical AI team, Margaret Mitchell.

This comes after weeks of being locked out of her work accounts over an investigation related to Mitchell’s objections concerning the controversial firing of her fellow co-lead Timnit Gebru.

According to a Google spokesperson, the investigation into Mitchell concerned alleged sharing of internal company files:

Our security systems automatically lock an employee’s corporate account when they detect that the account is at risk of compromise due to credential problems or when an automated rule involving the handling of sensitive data has been triggered. In this instance, yesterday our systems detected that an account had exfiltrated thousands of files and shared them with multiple external accounts. We explained this to the employee earlier today.

The firing of Timnit Gebru sent shockwaves throughout the AI community. It’s been widely viewed as a move to remove voices of dissent when those voices, world renowned ethicists hired specifically to investigate and oversee the ethical development and deployment of Google’s AI systems, don’t say what the company wants to hear.

Details are still coming in, but it appears as though Mitchell’s been let go as the result of Google’s investigation.

This story is developing…

Published February 19, 2021 — 22:26 UTC

Repost: Original Source and Author Link

Tech News

Nothing revealed by OnePlus founder Carl Pei

There’s a company called Nothing in the works, courtesy of Carl Pei. This is one of the people that made OnePlus happen a few years ago – he left OnePlus recently. Now, Carl Pei’s aiming to make something interesting. The company’s name, Nothing, works with the description of said company and all the details offered up thus far. It’s all intentionally vague and intriguing.

Carl Pei released Nothing today as a new venture, based in London, with a description that reads as follows. “A new, forward-thinking consumer technology company.” UPDATE: More details in the mix!

As the video released today suggests, “It’s easy to make something. Even easier when it’s just like the thing before it. And the one before that. But like all good things, this one starts from scratch. No notes. No blueprints. No map to find our way back.”

The intro video for this company goes on to say, “We’re rethinking everything, from what we make and how we make it… to what goes in and what goes out.*”

*At this point in the video, we see a couple of measurements. One says SPACE 2.5mm, the other says SPACE 5x5mm. We COULD assume this means they’re making a processor chip – that’d work with the 5x5mm, but a processor chip that’s also 2.5mm thick is… pretty thick.

“A giant reset button for all things innovation,” says the intro video for Nothing (aka @nothingtech on Twitter). “And so we go, confident that what’s in reach isn’t worth reaching for. We know, because we tried reaching a little further, and came up with… NOTHING.”

This release also suggests that “technology should fade into the background and feel like nothing.” The Nothing dot Tech website is ready to roll now. We’re in the process of getting a whole lot more information about this company – and what it’ll make – so stay tuned!

Repost: Original Source and Author Link