DATA FIDELIS SERVICES

HONG KONG (SAR) • C++17 / C++20 / C++23 HIGH-PERFORMANCE SYSTEMS FOR HFT & INSURANCE
BUSINESS ACTIVITIES (SOFTWARE SERVICES ONLY)

We build systems that survive market open, not just compile. Quant infrastructure — hardened for Trading Desks and HK insurers. Direct principal engagement from design to production cutover, with MSA-ready delivery, audit logs, runbooks, and on-prem / VPC deployment options.

PRODUCTION FOCUS

C++17, C++20 and beyond — High-Performance Systems for HFT & Insurance

Modern C++ from C++17 to C++20 / C++23: concepts, coroutines-ready, constexpr, lock-free, PMR allocators, zero-deps.
FEATURED METRIC
0.40µs
p99 HKEX OMD-C
2.04M/s
ticks/sec • C++20
SELECTED SYSTEMS — ALL PUBLIC GITHUB
• 4 PUBLIC repos • MIT • Open Source • v24 complete
01 • hkex-omdc-fabric.cpp — HKEX OMD-C Market Data Fabric (NEW) • C++20 / C++23 • TOP
PUBLIC • p99 0.40us • 2.04M ticks/sec • C++20 • PROD

HKEX OMD-C Low-Latency Fabric — Greenfield, Open Source

Production-grade UDP multicast fabric for HKEX OMD-C 239.1.1.1:5000. Built with C++20 concepts, SPSC lock-free ring, AF_XDP socket, cache-line padding. Open source for reviewable, extensible desk-specific alphas — validates at p99 0.40us on commodity Latitude 7480, 10× faster than typical 5us HFT target. MSA-ready with audit logs, runbooks, deterministic replay, MIT licensed.

PERFORMANCE — LATITUDE 7480, GNU 13.3.0, 1M TICKS, C++20 -O3 -FLTObench_omdc_fabric
P50
0.17us
P90
0.20us
P99
0.40us
P99.9
0.91us
MAX
51.62us
THROUGHPUT
2038k/s
GITHUB — PUBLIC • MIT • C++20/23
github.com/nsmanju/hkex-omdc-fabric
Build: cmake -O3 -flto -march=native
Stack: C++20 • SPSC • AF_XDP • cache-aligned • PMR
License: MIT • Public
02 • hk_claimsentry.cpp — Insurance Claims Integrity • C++17 / C++20 • PUBLIC
PUBLIC • 500K/sec • ML • PROD • MIT

HK-ClaimSentry 500K/sec — Real-time claims fraud & duplicate detection

Regex + embeddings hybrid pipeline with ONNX inference, RocksDB dedup store, Kafka streaming, and SFC Type 9 MSA audit trail. p99 < 8ms at 500K claims/sec. Designed for HK insurers requiring explainable fraud rules + ML hybrid with deterministic replay, runbooks, and on-prem/VPC deploy. Open source, MIT, production-hardened for market open survival.

GITHUB — PUBLIC • MIT • OPEN SOURCE
github.com/nsmanju/hk-claimsentry
SFC Type 9 MSA audit ready
500K claims/sec • p99 <8ms • PUBLIC
03 • fix_engine_equities.cpp — FIX 4.4 Equities Low-Latency Engine (NEW) • C++20 • PUBLIC • v4
PUBLIC • p99 <50us • 1.5B msg/sec • C++20 • MIT

FIX 4.4 Equities — QuickFIX p99 Fix, Zero-Copy, Ubuntu 24.04 CLI

Problem: QuickFIX/J is tool allocating std::string per field → 200-500us p99 breaks sub-2ms SLO for AAPL/MSFT/HSBA.HK.
Solution: C++20 zero-copy — buffer IS FixBuilder.build() = recv(fd,buf) wire buffer. FieldView {tag, string_view} slices original buffer, fields_[128] stack, linear scan '=' & SOH 0x01, auto 9= & 10=. Latitude 7480 validated, pedantic clean, e6384a7.

PERFORMANCE — LATITUDE 7480, 1M PARSES, C++20 -O3 -march=native -Wpedantic./fix_demo
P50
~5.2us
P90
~8.1us
P99
<50us
CYCLES
12632
BENCH
665us/1M
THROUGHPUT
1.5B/s
GITHUB — PUBLIC • MIT • C++20 • NEW
github.com/nsmanju/fix-engine-equities
Build: make -O3 -march=native -Wpedantic
Commit e6384a7 v4 final clean • equities only • no crypto
Next: fix-engine-equities-prod epoll session recovery
Open Repo — v4
04 • hft_straddle.cpp — Straddle Mispricing Engine • C++17 / C++20 • PUBLIC
PUBLIC • 1.2M ticks/sec • HFT • MIT • FPGA-ready

HFT Straddle — Volatility surface & gamma scalping

Sub-ms order book reconstruction, IV surface arbitrage, real-time gamma calculation with lock-free queues and kernel bypass. < 1.2ms p99 tick-to-signal. Built for prop desks & family offices running volatility arbitrage with FPGA-ready hooks, zero-deps deploy, and deterministic replay. Open source, MIT, C++17 to C++20/23, cache-aware, NUMA-aware.

stack: C++17 / C++20 / C++23 • FPGA-ready • lock-free queues • kernel bypass • IV surface • gamma scalping • MIT • PUBLIC

GITHUB — PUBLIC • MIT • OPEN SOURCE
github.com/nsmanju/hft-straddle-engine
Prop desks & family offices
Sub-ms recon • <1.2ms p99 • PUBLIC
CHIEF CONSULTANT

MANJUNATH, Nadkalpur Shivabasavanna

ORG
Data Fidelis Services
LOC
Hong Kong (SAR)
EMAIL
billing@datafidelis.com
PHONE
+852 5783 0107, WhatsApp

20y+ market systems, trading infra, HKEX connectivity, SFC MSA, low-latency C++17/C++20/C++23. Direct principal — no bench, no handoff. All repos PUBLIC, MIT, open source.

WHY DATA FIDELIS

We build systems that survive market open, not just compile. Quant infra — hardened for Trading Desks and HK insurers. All public, reviewable.

FOCUS
C++17, C++20 and beyond — High-Performance Systems for HFT & Insurance — All PUBLIC
© Data Fidelis Services • Hong Kong (SAR) • billing@datafidelis.com • +852 5783 0107 • security.txt • 4 PUBLIC REPOS
• PUBLIC • HK • C++20/23 • MIT