Streaming platform engineers | Video infrastructure architects | Teams operating encoding workflows at scale
THANK YOU FOR JOINIG US IN STOCKHOLM
Thank you to everyone who joined NETINT and Scalstrm in Stockholm for our Streaming Architecture & Encoding Efficiency event. This was our first event in this format, and we see it as a success. The conversations were practical, technical, and focused on real infrastructure challenges facing video engineering teams today, from encoding efficiency and workload scaling to where hardware acceleration fits into modern streaming architectures.
A special thank you to everyone who participated, asked questions, joined the demos, and helped make the discussion useful for the room.
Watch the Afternoon Technical Sessions
If you missed the event, or if you would like to revisit the technical discussions, you can watch the recorded afternoon sessions below. These sessions cover practical technical topics from the afternoon program, including real-world encoding workflows, infrastructure considerations, and hands-on discussions around VPU-based acceleration.
How Video Encoding Actually Runs: CPU vs Dedicated Hardware
Craig Butlin explains how video encoding works under the hood, from input and demuxing to decoding, preprocessing, scaling, and final encode. He compares CPU-based encoding with dedicated video hardware, showing why CPUs can struggle with sustained, high-throughput workloads and why VPUs offer more predictable performance, higher density, and better power efficiency at scale.
Unlocking True Throughput with NETINT VPUs in Live Streaming Workflows
Dominique Vosters explains how SCALSTRM uses NETINT VPUs inside a broader live streaming workflow that goes beyond basic transcoding. The session covers how VPU acceleration fits with CPU-side processing, frame repair, synchronization, packaging, audio handling, redundancy, and API-based management for carrier-grade live video delivery.
Integrating Hardware Encoding into an Existing FFmpeg Pipeline
Craig Butlin walks through a practical FFmpeg demo comparing software encoding with NETINT hardware encoding. He shows what actually changes in the pipeline, including decoder and encoder selection, hardware frames, filtering, scaling, and encode parameters, while demonstrating the performance difference between CPU-based and VPU-based transcoding.
Interested in the next Stockholm Event (May 2027)?
We are considering hosting another small technical event in Stockholm next year, directly after Streaming Tech Sweden.
If you work in video engineering, streaming infrastructure, cloud video, CDN workflows, encoding, transcoding, or media delivery, we’d love to hear from you.
Please fill out the form below to express interest in participating in the next event.
Space will likely be limited again, so this will help us plan the right format, topics, and technical focus for next year.
2026 GATHERING: STREAMING ARCHITECTURE:
ENCODING EFFICIENCY AT SCALE
Understanding how encoding decisions impact cost, scalability, and system behavior in production.
An independent technical gathering for streaming engineers and platform teams
MAY 22, 2026 // HOTEL AT SIX, MUSIC LOUNGE // STOCKHOLM
Taking place the day after Streaming Tech Sweden. Independently organized by NETINT and Scalstrm.
FOR: Streaming platform engineers | Video infrastructure architects | Teams operating encoding workflows at scale
IN STOCKHOLM FOR
?
If you are attending Streaming Tech Sweden (STSWE), join NETINT and Scalstrm for a focused technical session designed for engineers and platform teams working on streaming infrastructure.
The agenda is built around practical implementation: how encoding systems behave in production, where inefficiencies and cost drivers emerge, and how architectural decisions translate into cost, scalability, and operational outcomes.
A key focus will be the practical adoption of next-generation technologies such as VPUs. How they are deployed, integrated into existing pipelines, and used to significantly reduce encoding cost and infrastructure footprint. Through technical sessions and peer discussions, we’ll examine real deployment models, performance trade-offs, and lessons learned from operating large-scale encoding systems in production.
AGENDA
FRIDAY, May 22 | 9:00 - 12:30 | Educational sessions
09:00 – 09:30 | Breakfast & Arrival
Light breakfast and coffee will be served during arrival.
09:30 – 09:40 | Leo Nieto, NETINT | Opening: From streaming architecture to encoding efficiency at scale
The streaming community is already discussing delivery, playback, cloud workflows, and emerging protocols during the week of Streaming Tech Sweden.
This opening session narrows the focus to encoding: how it fits into streaming systems and why it is becoming a primary factor in cost, scalability, and infrastructure design.
By connecting encoding decisions to broader architectural outcomes, this session sets the stage for a deeper, system-level conversation throughout the morning.
09:40 – 10:10 | Leo Nieto, NETINT | Encoding is still the most expensive line item in streaming
As streaming workflows scale, encoding quietly becomes the dominant cost driver, often without visibility.
This session breaks down where those costs come from, why software-based approaches struggle at scale, and how infrastructure decisions made early in pipeline design can lead to 5-10x cost differences.
The goal is not to introduce a product, but to challenge how encoding is currently approached across the industry.
10:40 – 11:00 | Coffee Break
Coffee, tea, and water will be available during the break.
11:00 – 11:30 | Dominique Vosters, Scalstrm | Where streaming infrastructure is actually heading
Streaming workflows are becoming more dynamic, distributed, and cost-sensitive. As content volumes grow and workloads become less predictable, static pipelines are no longer sufficient.
This session explores the broader shift toward orchestrated, resource-aware infrastructure, highlighting the operational and economic challenges teams face today, and what needs to change to support the next generation of streaming at scale.
11:30 – 12:00 | Pontus Eklöf, Scalstrm | From pipelines to systems: orchestrating encoding at scale
Managing encoding at scale is about coordinating resources, workloads, and priorities across an entire system.
This session dives into how modern orchestration approaches enable better utilization, automation, and cost control. Through real-world examples, we’ll explore how workflows can be designed to adapt dynamically to demand while maintaining efficiency and reliability.
12:00 – 12:30 | All speakers | Panel Discussion: What does a cost-efficient streaming stack actually look like in 2026?
With increasing pressure on cost, scalability, and performance, how should modern streaming architectures be designed?
This discussion brings together perspectives from infrastructure, encoding, and orchestration to explore trade-offs between CPU, GPU, and dedicated hardware, cloud versus hybrid models, and how teams can balance flexibility with efficiency. Expect an open, technical conversation grounded in real-world experience.
12:30 – 13:30 | Lunch
A light lunch will be served, including a lunch bowl, bread, coffee, and a sweet option.
Hands on Sessions
Two parallel tracks: technical deep dives and 1:1 architecture clinics
FRIDAY, May 22 | 13:45 – 15:00 | Technical Deep Dives (Engineering Track)
Short, practical sessions focused on real implementation challenges in encoding and workflow orchestration.
Each session breaks down how things work in practice – from integration to system behavior – so you can understand what it would take to apply these approaches in your own pipeline.
Attendees can join individual sessions or stay for the full sequence, followed by an open Q&A.
13:45–14:00 | Session 1 | Craig Butlin, NETINT | How encoding actually runs: CPU vs dedicated hardware
What actually happens inside an encoding job, and how does that change when you move beyond CPU-based workflows?
This session breaks down the encoding process step by step, from decode to output, and examines where traditional CPU-based approaches begin to hit limits at scale. It then explores how dedicated encoding hardware changes execution models, performance characteristics, and system behavior.
Rather than focusing on theory, the session provides a practical view into how encoding runs in real environments, what changes when new compute is introduced, and what remains consistent across workflows.
What you’ll learn:
- How an encoding job actually executes (decode → process → encode → output)
- Where CPU-based encoding struggles at scale
- How dedicated hardware changes parallelism and throughput
- What stays the same (FFmpeg workflows, pipelines)
- What changes in performance, density, and expectations
Takeaway:
A clear, practical understanding of how encoding works under the hood, and what changes when you move from CPU to dedicated hardware.
14:00–14:15 | Session 2 | Dominique Vosters, Scalstrm | What actually happens when you orchestrate encoding jobs
Orchestration is often described at a high level, but what actually happens inside the system when encoding jobs are scheduled and executed?
This session breaks down how orchestration works in practice, from job queueing and scheduling decisions to how systems determine where and when workloads run. By comparing orchestrated and non-orchestrated pipelines, it highlights where inefficiencies arise and how different approaches impact performance, cost, and scalability.
Through real examples, the session shows how the same workload can behave very differently depending on orchestration logic, and what needs to be visible and controllable to run encoding systems effectively.
What you’ll learn:
- How encoding workflows operate without orchestration
- Where inefficiencies typically occur in static pipelines
- How job queueing and scheduling decisions are made
- How systems determine where and when jobs run
- How different orchestration strategies impact the same workload
- What you need to monitor and control (observability)
Takeaway:
A practical understanding of orchestration as decision logic, and how it directly shapes system efficiency and behavior at scale.
14:15–14:30 | Session 3 | Craig Butlin | NETINT | Integrating hardware encoding into an existing pipeline
What does it actually take to introduce dedicated encoding hardware into an existing workflow?
This session walks through how hardware encoding integrates into typical FFmpeg-based pipelines, focusing on what changes in practice. From command-level adjustments to deployment models and system behavior. It covers how to incorporate new compute without redesigning your entire pipeline, and what to expect when moving from CPU-based encoding to hardware-accelerated workflows.
With a focus on real implementation, the session also highlights common pitfalls, including scaling assumptions, I/O constraints, and debugging challenges.
What you’ll learn:
- How hardware encoding fits into a typical FFmpeg-based pipeline
- What actually changes (commands, flags, device selection)
- Integration options: bare metal, containers, and Kubernetes
- How to handle concurrency and multiple encoding jobs
- Common pitfalls, including scaling assumptions and I/O bottlenecks
- What to monitor and how to validate output
Takeaway:
A practical understanding of how to integrate hardware encoding into existing workflows without rebuilding your pipeline from scratch.
14:30–14:45 | Session 4 | Dominique Vosters | Scalstrm | Building a pipeline that adapts instead of breaking
What happens when encoding pipelines are pushed beyond their limits, and how do you design systems that can handle real-world variability?
This session explores why traditional, fixed-capacity pipelines struggle at scale, particularly under uneven workloads and traffic spikes. It then introduces the principles and design patterns behind adaptive systems, including queue-based architectures, workload separation, and priority-driven processing.
By looking at how modern pipelines are built to respond dynamically to changing demand, the session also covers hybrid environments and the balance between automation and control.
What you’ll learn:
- What typically breaks first in encoding pipelines at scale
- Why static, fixed-capacity systems fail under real workloads
- Key design patterns for adaptive pipelines (queues, prioritization, workload separation)
- How to handle spikes and uneven demand effectively
- How hybrid setups (cloud + on-prem) fit into scalable architectures
- Where to draw the line between automation and manual control
Takeaway:
A practical framework for designing encoding pipelines that adapt to real-world conditions—and continue to perform reliably at scale.
14:45–15:00 | Session 5 | Dominique Vosters, Scalstrm & Craig Butlin, NETINT | Bring your pipeline - Architecture teardown (whiteboard session)
An open, technical discussion focused on real-world streaming architectures.
In this interactive session, participants are invited to bring their current pipeline, challenges, or constraints. Together, we’ll walk through real setups, discuss trade-offs, and explore how different architectural choices impact cost, scalability, and performance.
Rather than theory, the session is driven by actual systems, providing practical insights into how pipelines behave in production and how they can be improved.
What you’ll get:
- Direct feedback on real-world pipeline architectures
- Discussion of trade-offs across compute, orchestration, and infrastructure
- Practical recommendations based on real constraints
- Exposure to how other teams are solving similar challenges
Takeaway:
A clearer understanding of how your architecture performs today—and what you can do to improve it.
Optional: Get your pipeline reviewed
Participants who would like their architecture discussed are invited to submit a short overview in advance.
Submissions are reviewed ahead of the session to ensure focused, high-value discussions.
A limited number will be selected for live review.
FRIDAY, May 22 | 13:45 – 15:00 | Architecture Clinics
(Business Track)
Short, focused working sessions, focused on identifying where cost and inefficiency are coming from, and what to do about it.
Availability
Sessions are limited and pre-booked to ensure each discussion remains focused and valuable.
How it works
Each 15-minute session follows a simple structure:
1. Pipeline snapshot (5 min)
You provide a brief overview of your current setup, including workload type (VOD, live, UGC), infrastructure (cloud, on-prem, or hybrid), encoding approach, key challenges related to cost, scaling, or complexity.
2. Rapid assessment (5 min)
We review your setup to identify where cost and inefficiencies are occurring, which parts of the pipeline are driving spend or limiting scale, how your approach compares to common system patterns, and what risks or trade-offs may arise when making changes.
3. Targeted recommendations (5 min)
We outline practical ways to improve cost efficiency and scalability, highlight architectural adjustments worth considering, and provide guidance on implementation risks along with the expected impact on performance and operations.
Speakers from Scalstrm and NETINT
Engineers and architects working on real-world streaming systems
Business Development & Sales Director
With over 20 years of experience in Digital TV, Broadcasting, Networking, and Advertising, Dominique Vosters bridges the gap between complex engineering and business impact. As the Business Development & Sales Director at Scalstrm, he drives sustainable and efficient media delivery, optimizing streaming performance while reducing costs and energy consumption. His expertise spans solution design, technical sales, and large-scale deployments.
Business Development & Sales Director
Pontus Eklöf focuses on helping streaming platforms optimize video delivery for cost, scalability, and performance. With a background in cloud, media platforms, and cybersecurity, he brings experience translating complex infrastructure challenges into practical solutions.
Pontus has worked with large-scale streaming and cloud environments, supporting broadcasters and service providers in improving efficiency, reducing infrastructure overhead, and scaling video delivery systems in production.
Field Application Engineer
Craig Butlin is a Field Application Engineer at NETINT Technologies, focused on real-world deployment of video encoding and streaming workflows. With a background spanning AWS Elemental, Ericsson, Amagi, and Techex, he brings hands-on experience across cloud, broadcast, and large-scale media systems. Craig specializes in practical integration, debugging, and performance optimization, helping teams understand how encoding infrastructure behaves in production and how to scale efficiently.
Director of Market Development, EMEA
Leonardo Nieto represents NETINT across EMEA, working closely with streaming platforms, telecom operators, and infrastructure providers scaling video in power- and cost-constrained environments. With experience spanning applied transcoding, live streaming, and high-growth technology ventures, he focuses on the practical realities of deployment: density per rack, watts per stream, codec transitions, and cloud versus edge trade-offs.
VENUE
Hotel At Six
Hotel At Six | Brunkebergstorg 6, 111 51 Stockholm, Sweden
Hotel At Six is one of Stockholm’s premier event venues, situated at Brunkebergstorg in the heart of the city. Steps from T-Centralen and with easy access to Arlanda Airport, it offers a central setting for engineers, architects, and infrastructure teams to gather for focused technical discussions.
*** This event is independently organized by NETINT and Scalstrm. It is not organized by, affiliated with, or endorsed by Streaming Tech Sweden or Eyevinn.
VPU Product Brochures
