.jpg)
In our recent town hall, we decided to onboard a group of IoT and Data technology enthusiasts and have them face-to-face with our CTO, Deep Nayar, for a Q&A session on the newly introduced Flex83 Version 7, and the outcome is super-insightful.
Deep Nayar: Existing hyperscalers, while inherently powerful, claim to assist in developing custom platforms but impose significant infrastructure costs and vendor lock-in, offering little control over unit economics at scale and limited deployment flexibility. Although building from open source presents an attractive path for engineering teams, it becomes a non-starter due to the substantial cost and time investment required, particularly when proven Application Enablement Platforms like Flex83 are readily available with rich feature sets.
So, we devised a different software approach where we started looking at how we can slice and dice the software in a way that it becomes partly agnostic to the business riding on top of it.
At first, we aligned with the term "advanced application enablement platforms" where we take the software side and divide it so we can provide instant value to business solutions built on top. We then redefined our AIoT platform that can take any data in and operate on that data quickly and effortlessly. You can interact with it in three ways: through the UI for low-code/no-code interface, through APIs if you prefer integration, or through a CLI shell that we're currently developing.
The platform features multiple connectors for data ingestion, followed by transformation and predicate processing before storing in our three-tier data architecture: KSQL databases for streaming processes, Iceberg-like data stores for batch analytics, and Druid databases for OLAP analytics. We've essentially made the entire data plane problem click-based.
Deep Nayar: We have integrated in system click to build analytics like H2O.ai (auto ML) as well as custom full featured custom analytics pipelines. The platform supports different regression models and classification models that users can leverage to build their own models. The system walks you through the entire feature engineering chain - data ingestion, transformation, and in the intelligence layer, you go through data cleaning, feature engineering, and training.
You can work with both telemetry data and rich data for audio and video object recognition, as well as NLP-based workloads for audio recognition. The platform gives you big data handling and AI/ML handling in a very easy, click-to-build fashion, so you can start bringing in your data and experimenting with it right away and make data available to consumers via secure catalog API.
Deep Nayar: The platform can be deployed anywhere with a single click. We can run it on any hyperscaler like AWS, Azure, and Google, and we can even bring the footprint down to run on a decent laptop. We have infrastructure-as-code scripts for all public and private clouds to launch the platform.
We can also reduce the footprint to run on gateways that are 8GB or 16GB in nature. It's a model where the Edge Brain - a smaller version of the platform with embedded agents - combines with the Data Brain and Asset Manager to help you onboard an asset in a connected world within hours of engagement. The multi-tenancy allows clients to use it in different ways, whether deploying on their own platforms or building applications on top for various environments like factory floors or private data centres.
Deep Nayar: We've engaged in similar partnerships before, where as a software provider, we recommend hardware to our clients when appropriate, and in return, are introduced to clients who need software solutions when they purchase hardware. Offering an integrated, ready-to-deploy solution benefits both companies.
We've developed northbound application solutions that partners can showcase to their clients, helping them envision the possibilities when working with specific hardware. This creates a model where hardware and software providers can deliver end-to-end solutions, reducing vendor complexity for customers who prefer single-source solutions rather than managing multiple providers.
Deep Nayar: We're currently adding auditable capabilities to the platform to ensure all changes are recorded in an immutable log. This feature was missing in our industrial IoT solution but is essential for compliance in healthcare and similar industries. GDPR compliance is already largely in place, covering privacy requirements, and the audit trail will be available by the end of this year to fully support healthcare market needs.
The healthcare market, particularly radiology, is very interested in on-premise solutions that offer complete privacy with low latency. We're focusing on privacy, anonymization, masking, and encryption for on-premise deployment. We should be able to demonstrate these industry-specific features by October, with a beta version targeted for the end of the year.
Deep Nayar: Recently, we’ve worked with an electronics test and measurement equipment manufacturer on low-latency signal processing where they needed sub-second or 300-millisecond processing for millions of data points in their signal analysis workflows. They specialize in hardware and very high-frequency bandwidth management for signal processing, dealing with millions of data points plotted onto local systems with data coming from different hardware for signal analysis.
We also work with companies like Renesas on model training implementations. Our clients typically work with different scales - either millions of streams that need to be processed in runtime, millions of endpoints managed through predictive maintenance, or millions of points plotted for graphing with nanosecond-worth collection frequency. The software is built to handle runtime batch processing and runtime analytics at very low latencies.
Deep Nayar: The trend is definitely toward integrated solutions. Customers care about the software outcome but don't want to get hands-on with on-premise hardware solutions. By partnering with hardware providers, we can offer bundled solutions that address this need.
We're seeing strong interest in Edge AI solutions where hardware and software work together seamlessly. The key is making the entire solution portable and cloud-agnostic, so it provides flexibility for teams to deploy in various environments - whether that's factory floors, public or private data centers, or client infrastructure.
The integration of LLMs - we work with Claude, Gemini, GPT-4, and GPT-5 depending on client preferences - along with traditional AI/ML capabilities, creates a comprehensive platform that can adapt to various business requirements while maintaining the performance and compliance standards that enterprise customers demand.
Deep Nayar: Look for platforms that separate the plumbing work from the business logic. You want something that gives you the foundational capabilities - the 80% that's common across applications - so you can focus your resources on the 20% that truly differentiates your business. The platform should provide instant value while allowing you to retain full IP ownership and control over your unique innovations.
Also, ensure the platform can scale with your needs, whether you're processing telemetry data, handling compliance requirements, or deploying across various environments. The real value comes from being able to get to market quickly while maintaining the flexibility to adapt as your business grows.
This townhall has been condensed and edited for clarity but preserves the richness of the original conversation between the group and Deep Nayar for the IoT Leadership Lounge series.
