Forem: Extrieve Technologies

Extrieve’s new version of Flutter SDK — Now with AI-powered, secure and regulation compliant offline document processing

Extrieve Technologies — Tue, 22 Jul 2025 09:48:02 +0000

A while back, we were working on a mobile app that needed KYC verification. But the minute we got into it, it turned into a mess — multiple SDKs, multi - API calls, security -challenged, managing diverse document formats, ensuring accurate document cropping, complex data (Like Aadhaar) masking rules and poor internet in the field. It’s a workflow that frequently introduces friction, latency, and compliance risks.

We wanted something simple: scan a document, extract the data, and keep it all offline. But nothing out there really ticked all the boxes. So, we ended up building it ourselves.

What came out of it is a Flutter plugin available now on pub.dev that wraps around our SplicerAi SDK. Here is how it transforms your document processing workflow

Full On-Device Processing: handles the full document processing flow (like KYC flow) on-device — no servers, no external API calls. You scan the document (Aadhaar, PAN, passport, etc.), and the plugin handles detection, extraction and validation

Automated Aadhaar Masking: It automatically hides everything except the last 4 digits of Aadhar, ensuring privacy and regulatory adherence.

AI/ML Baked-In: The SDK has built in AI/ML models and it accurately handles various ID documents (Aadhaar, PAN, Passports, Driving Licenses, Voter IDs, etc.)

Robust Offline Capability: This is a game-changer. For applications in remote banking, insurance field onboarding, logistics, or government services where internet access can be unreliable, our SDK ensures seamless operations. Imagine instant KYC verification for agents in rural areas or during power outages – no more dropped transactions or frustrated users.

Extrieve made it developer-friendly too. Add the dependency, grant the required permissions, init the SDK, and you’re good to go. Works for both Android and iOS via Flutter.

You can check it out here:
Introducing the Flutter Plugin for SplicerAi SDK by Extrieve

We built this for ourselves originally but figured it might help others deal with the same pain. If you're working on something similar, give it a try — and feel free to reach out. I would love to hear how you’re solving document processing in your apps.

Trying to Run AI on Low-End Devices? Here’s What We Learned

Extrieve Technologies — Thu, 17 Jul 2025 10:03:16 +0000

We recently built something that frankly we weren’t sure was possible when we started: a lightweight AI engine that runs document detection and face matching directly on the edge — on phones, desktops and even browsers, completely offline. And we didn’t use TensorFlow, ONNX or any of the usual ML frameworks. The idea was to make it run anywhere, without internet or GPU dependency, and actually perform well.

We ran into all kinds of issues — bloated model sizes, slow browser load times, memory limits on mobile and the usual cross-platform headaches. So instead of patching together existing tools, we ended up writing our own inference engine in C++, keeping it minimal, and wrapping it for Android (NDK), iOS, Flutter, React Native, WebAssembly and desktop. For the models, we trained them to stay under 15MB and optimized them to be fast and resource-efficient. One of those models KIMORA is our document understanding model — it can detect corners, fix skew and understand layout even in tricky conditions like poor lighting.

The toughest part was getting it to run smoothly in browsers. We chunked the model and binary files, downloaded them in parallel, and cached them for instant reuse — this alone brought our browser AI load time down. Looking back, building everything from scratch gave us more control, but it also taught us how hard real edge AI is when you actually try to make it work everywhere.

If you’re working on lightweight AI for constrained environments, or just tired of relying on bloated ML stacks, I hope this gives you a few useful ideas.

If you’re curious to know more ([https://blog.extrieve.com/general/from-idea-to-execution-building-a-cross-platform-ai-engine-for-the-edge/]

Would love to hear your thoughts and happy to answer any questions if you’re working on something similar!

Workflow Automation Fails When You Try to Remove the Humans

Extrieve Technologies — Tue, 15 Jul 2025 11:28:48 +0000

There’s a lot of enthusiasm around automation. We all want to move faster, reduce manual steps, and make systems more predictable. But after working on several workflow projects — both as builders and as partners to large organizations — I’ve come to a pretty grounded conclusion:
Most workflow automation systems fail because they try to automate too much.

At first glance, automation looks like a clean fix. You define the process, build the flows, apply the rules, and let the system run. But real life never sticks to a clean flowchart. Input isn’t always perfect. Scenarios don’t follow scripts. And people — with all their judgment, flexibility, and ability to work through ambiguity — aren’t something you can just remove.

In almost every project we’ve seen go off-track, the failure point wasn’t the tech stack. It was the assumption that everything could be handled automatically. And when something didn’t fit the model, users were forced to handle it outside the system — by email, by memory, or by writing things down. That’s where visibility breaks. That’s when trust in the system starts to fall apart.

We decided to approach this differently.
Instead of treating human involvement as a failure case, we built our system — PowerFlow — to expect it. The idea is simple: automation should handle what it can, and humans should step in when needed — within the workflow, not around it.
That means:
When an AI agent can’t confidently verify something, it routes the case to a human reviewer.
When an exception occurs, it doesn’t stop the process — it adapts.

Every manual decision is logged, tracked, and visible across the case lifecycle.
And just as important, we made sure the system itself is flexible. Teams can define their own queues, routing rules, and even field-level logic — without constantly pulling in developers. In one setup, we had a KYC document process where the AI would read uploaded IDs, extract names and photos, and validate them. If anything didn’t match, it would go to an operations team member, who would see exactly why the case was flagged — and could resolve it within the same system.

The final outcome (whether automated or reviewed) was always stored with a clear trail of who did what, when, and why.
This kind of design — where human-in-the-loop isn’t a fallback but a planned, visible part of the system — has made a huge difference in adoption and reliability.
If you’ve ever rolled out workflow tools that quietly fall back to manual steps, or if your automation pipelines break down on bad input, you might know exactly what I’m talking about.

We wrote a more detailed breakdown of this problem (and how we handled it) here:
** Why Enterprise Workflow Automation Fails — and How to Do It Right**

I’d love to hear how others are handling similar challenges.