SoftFlame Ventures LLC is proud to congratulate our Lead Technical Service Specialist, Md Junaeid Alam, and his team on publishing their first AI research, a part of AI based HR Platform Studio Software
Here's what it does and why it matters
Imagine you manage 500 employees across multiple offices. Every day, your mobile attendance app logs thousands of check-ins. But how do you know who's actually showing up versus who's letting a friend scan for them, faking their GPS location, or quietly slipping out 30 minutes early every single day?
That's the problem Md Junaeid Alam set out to solve as part of the AI HR Platform Studio Software , a flagship initiative under SoftFlame Ventures LLC
Most attendance fraud systems give you a binary answer: fraud or not fraud. That's not how the real world works. Some employees are borderline is worth a second look, but not worth blocking. So they built an AI model that thinks in three tiers:
- ✅ Normal → Auto-approve and move on
- ⚠️ Flag → Send to AI HR Studio Software for review
- 🚫 Fraud → Block immediately and alert
The model is a hybrid CNN-BiLSTM-Attention architecture trained with Supervised Contrastive Learning , a technique that teaches the model not just where the decision boundary is, but how to shape the entire embedding space so borderline cases become visible.
The outcome?
- ✅ 100% fraud recall (zero missed fraud cases)
- → 98.63% accuracy
- → 0.23 MB model size — small enough to run on-device in under 10ms
- it's the foundation of something much bigger we are building together AI based HR Platform Studio Software.
We couldn't be prouder of Md Junaeid Alam and the entire team for this milestone. This is what happens when vision meets execution
📄 Read the full article on SSRN (Report No. 6649058):
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6649058
Stay connected as we roll out our suite of products and services, and join us on this exciting journey where technology meets ambition.