Company
Altruist is the tech-forward custodian for independent RIAs, offering a fully integrated digital platform that makes managing investments and serving clients simpler and more affordable. Altruist combines a self-clearing brokerage firm with intuitive software for account opening, trading, reporting, and billing. With Altruist, financial advisors can create custom portfolios, trade fractional shares, automate rebalancing, and provide clients with a sleek web and mobile app experience.
Challenge
Altruist’s platform consumed significant cloud resources for daily workloads, including end-of-day processes that triggered sharp, short-lived surges in demand, continuous reporting, and performance monitoring environments running 3-5x normal production traffic. To keep it cost-efficient, workloads were divided across working groups. Managing this complexity while controlling cost and scalability was a challenge.
Solution
Cast AI makes it easy for the Altruist team to balance resources, while compute instance provisioning and cost projections became core tools in daily operations.
Within 60 to 90 days, the team started seeing consistent savings of 45% or more, with even greater reductions during less busy times of the week. Cast also prevented downtime by automatically scaling pods during unexpected production issues, proving its reliability. With proactive scaling that adds or removes capacity on demand, teams avoid both overpaying and downtime, keeping cloud environments perfectly balanced at all times.
Moreover, automation frees 85–130 hours every month that engineers once spent on manual tasks and incident response, reducing operational overhead but also delivering lower costs and faster response times across all environments.
Results
- 45%+ cost savings
- Engineers save 85–130 hours per month previously spent on manual work and incident response
- Every $1 spent on Cast AI delivers $13–$26 in value
Dramatic cost reduction thanks to Cast’s automation features
Cast’s automation features reduced cluster costs across Altruist clusters, as illustrated by the downward trend in the image.
Increase in CPU and memory utilization
Workload Autoscaler intelligently adjusts workload requests to guarantee optimal resource utilization for Altruist’s workloads.
In the image above, the dark blue line indicates resources allocated for a certain workload, the light blue field shows resources requested, and the green one represents resources that were really used – all of which show the level of overprovisioning Cast helped to decrease from a specific point during one month.
Overprovisioning levels before and after integrating Cast
Spot Instance automation
Cast automates the entire Spot Instance lifecycle for Altruist, allowing the team to increase the share of these cost-efficient compute resources across their clusters.
One of the things that attracted us to Cast AI was its automated approach: providing data on how much you can save and projecting possible cost savings. Cast AI not only gave us a north star to see where we could improve, but also helped us be proactive in managing workloads in that way.
Cast unlocked a lot of our time for experimentation and innovation rather than daily toil. There’s a bit of “magic,” if you want to call it that, in the way Cast works – and my team enjoys that.
Emerson Serra, Sr Director Cloud, Security & AI Engineering at Altruist
Running a high-performing and cost-efficient platform
What requirements does your infrastructure need to meet to support your financial services platform?
Because of the nature of our platform, we need to ensure it’s reliable and provides the performance our customers need. We always keep security in mind and operate in a very secure environment.
We have spiky end-of-day and start-of-day processes. We also generate many reports throughout the day to ensure our customers always have the latest information about their transactions. At the same time, we run internal environments with continuous performance monitoring, where we handle three to five times the amount of tracking we would normally receive in production.
What’s important is that we run some workloads within specific working groups; that approach allows us to ensure that our applications stay balanced.
To ensure everything runs smoothly, we’re also very conscious about cost and scalability at all times. That’s where Cast AI comes into play for us.
How did you approach cost management and optimization across your applications?
We were pretty much set in the traditional AWS Reserved Instances world. As you know, that can be daunting – you need to stay on top of things and have the right monitoring in place.
One of the things that attracted us to Cast AI was its automated approach: providing data on how much you can save and projecting possible cost savings. Cast AI not only gave us a north star to see where we could improve, but also helped us be proactive in managing workloads in that way.
Emerson Serra, Sr Director Cloud, Security & AI Engineering at Altruist
Have you considered open-source solutions for optimizing cloud costs?
I used the open-source tool from Kubecost in the past, and we implemented it at some point. In previous roles, I always had Kubecost in some capacity. But for obvious reasons, at the enterprise level, I pivoted to Cast.
A move towards automation
What was onboarding Cast like?
I’d say it was fairly straightforward; Cast is designed to be easy to install.
Emerson Serra, Sr Director Cloud, Security & AI Engineering at Altruist
One of the lessons we learned was ensuring all our applications were stateless and identifying which workloads needed to be in a different workgroup. For instance, every day, we run builds, tests, and other processes that sometimes require a different approach. We can’t put those on agile Spot Instances. Cast made it easy to move workloads in that way.
At the moment, we’re only using Cast in non-production environments, where our cloud footprint is actually bigger than in production. The reason we don’t yet have it in prod comes from a combination of factors, like internal processes that we need to mature to build that confidence. Right now, Cast AI provides an analysis of what’s happening in the cluster at a certain capacity, but we still have work to do on our side before we can fully graduate to production.
In the two or three years we’ve been working together, there has been no downtime, so we’re confident about our journey.
Emerson Serra, Sr Director Cloud, Security & AI Engineering at Altruist
What is the support like?
We’re happy with the partnership. There’s a lot of proactiveness, and I feel like we’ve also seen the Cast product grow so much, which is great. We were exchanging messages about a drop even today, so the Cast team has been very active.
What level of cost savings did you achieve with Cast?
I would say it’s always around 45% or more. Sometimes, depending on the time of the week, we see more dramatic cost savings. Toward the end of the week, it’s not as busy since we stop our continuous testing on Friday, so the drop in infrastructure costs is more noticeable.
We also see it after post-release checks are completed, since our release cycle usually happens in the middle of the week. Overall, the savings have been fairly consistent at 45% or higher.
What was your ROI timeline?
It was pretty fast, to be honest. We were very aggressive savings-wise when we started with Cast AI. We had been looking at different open-source solutions and trying to manage schedules ourselves, and then we found Cast. At the same time, we started migrating some clusters to more modern, better-designed platforms on our side. I’d say we started seeing results in about 60 to 90 days.
We use Cast in non-production environments, but as I mentioned, it’s not enabled in production. So we don’t see those benefits there yet. Still, it was a very good eye-opener and showed us that enabling it in production, even at some capacity, could help with stability.
Our return on investment with Cast AI has been exceptional, ranging from 1,350% to 2,600%. In practical terms, that means every $1 we spend delivers $13–$26 in value through automation-driven cost savings, efficiency gains, and reduced engineering overhead.
Emerson Serra, Sr Director Cloud, Security & AI Engineering at Altruist
Automation turned out to be a game-changer
Which Cast features were real game-changers for you?
Compute instance selection and autoscaling
EC2 provisioning is probably the one we rely on the most. As I mentioned, we also pay close attention to the cost projections. Managing groups is pretty straightforward, and that’s another area we take advantage of. But for now, instance provisioning within our Kubernetes clusters is our bread and butter.
Workload Autoscaler
Cast has a feature that compacts workloads; it looks at the instances and makes sure we’re utilizing them as much as possible. At one point, we were stable in our test environment with some applications under load, and then we got hit in production with a bug. Cast was able to take care of it without downtime, which was pretty eye-opening. It handled the situation by vertically scaling the pods, and we didn’t have that benefit in production before.
Did Cast improve engineer productivity in your team?
Cast AI has significantly improved our engineers’ productivity. We’ve freed up 85–130 hours every month (previously spent on troubleshooting, monitoring, and operational overhead) by automating manual work and incident response. Tasks like capacity management, performance tuning, and policy configuration now run automatically, allowing our team to focus on higher-value initiatives. These efficiency gains have not only reduced operational costs but also accelerated response times across all environments.
Emerson Serra, Sr Director Cloud, Security & AI Engineering at Altruist
Cast unlocked a lot of our time for experimentation and innovation rather than daily toil. There’s a bit of “magic,” if you want to call it that, in the way Cast AI works – and my team enjoys that.
What are the next steps you’re planning to take with Cast?
The features we haven’t adopted yet but are considering include the AI-driven model provisioning (AI Enabler) and Database Optimization. A couple of people on my team are testing the database optimization feature, since we already use a proxy that provides us visibility and integrates some AI. We’re seeing potential gains there.
