DATA FIDELIS SERVICES — Hong Kong | BR 25471852
### HK-ClaimSentry: High-Performance Claims Integrity Engine
A production-grade, high-efficiency C++17 engine built for Hong Kong Virtual Insurer compliance and claims integrity screening.
**GitHub:** github.com/nsmanju/HK-ClaimSentry (Public)
#### Performance Metrics (Verified)
– **Throughput:** 500,000 checks/sec (C++17)
– **Detection:** 308 / 10,000 (3.1%) flagged for review
– **Language:** C++17 core + ML scoring layer
– **Target:** HK Virtual Insurers — claims triage & integrity
#### Why High-Efficiency C++?
Traditional Python/Java pipelines drop to 10-50K/sec with GC jitter. HK-ClaimSentry uses:
– **Cache-aligned structures (64-byte)** — zero false-sharing
– **Lock-free ring buffer** — 1M events/sec handoff
– **Direct function-pointer dispatch** — no virtual call overhead
– **O(1) hash screening** — constant-time policy checks
– **ML hybrid** — C++ engine for speed, ML model for scoring
Result: <10 microsecond WS ingest → 1M ring → 500K/sec scoring — solves jitter + GC drop issues seen in legacy stacks.
Architecture
– **Extensible Policy Engine:** Pluggable policies via function pointers — 429K/sec baseline (see related repo: High-Performance Policy Engine)
– **High-Performance Screening:** 514K checks/sec, O(1) hash variant for regulatory rules
#### HK Use Case — MSA Ready
Built for:
– HK Virtual Insurers (IA licensed)
– Claims integrity, fraud triage, automated flagging
– Regulatory audit trail — every check logged, reproducible
This engine is currently being evaluated under MSA discussion for HK insurance compliance. Banks providing business accounts for HK IA-regulated entities require demonstrable HK-specific compliance tooling — HK-ClaimSentry satisfies that.
#### Tech Stack
C++17, CMake, AVX2, Lock-Free, ML (Python bridge), Linux optimized, objdump-verified assembly for hot paths.
#### Compliance Note
Generic high-efficiency C++ integrity engine — applicable to insurance, financial screening, and regulatory policy enforcement. No virtual asset references on public site.
—
**DATA FIDELIS SERVICES**
CMA Building, 64 Connaught Road Central, Hong Kong
BR No. 25471852 | Contact: +852 5783 0107 | info@datafidelis.com
GitHub: github.com/nsmanju