Dnalyaw

Algorithmic Trading
Infrastructure

OMS, risk engine, execution, and normalized market data. From backtest to shadow trading to live. Multi-market, multi-strategy.

Dnalyaw Trading System Architecture
varies
~100ms network
<1ms hot path
<10ms warm path
TWS ~100ms · FIX ~10ms
varies
Data VendorREST + Binary feed
IB TWSTCP · Streaming
Futu OpenDTCP protocol
Market Data
Provider InterfaceConnect · Subscribe · Stream
NormalizerSymbol · Quote · Trade
Quality GateSpread · Spike · Stale
Symbol ResolverAAPL → AAPL.NASDAQ
Trading EC2
Hot Path
Risk Engine8ns single check
Pre-trade ChecksExposure · Limits · Veto
Circuit BreakerKill Switch · 5% daily
Warm Path
OMSp99 7.8µs submit
Order State MachineLifecycle · Idempotency
ReconcilerPosition · Fill matching
Position LedgerT+1 settlement · Fills
Scheduled JobsStats 08:00 · TCA
Venues
IB TWSPrimary · Algos · Multi-asset
IB FIX / DMAP7 · ~10ms
AlpacaREST · Fallback
FutuTCP · HK paper
NASDAQ · NYSEUS Equities
HKEXHK Equities
PostgreSQLTimescaleDB
API EC2
Market APIREST · gRPC
BacktestREST API
Alerts · Bars · SSE Streaming
Cold Path
Quant ResearchBacktest · ML
Strategies6 built-in
Feature Store · Model Registry
Data flow
Order flow
Scheduled
Future (P7)

Four-gate risk veto.

Every order passes through exposure, position limit, sector cap, and daily loss checks. The engine approves, rejects, reduces, or flattens — with no override path.

Orders In
Risk Engine
Decisions
Exposure
Position limit
Sector cap
Daily loss
AAPLBUY
Exposure
AAPLAPPROVED
TSLASELL
Daily loss
TSLAREJECTED
NVDABUY
Position limit
NVDAREDUCED
METABUY
Sector cap
METAREJECTED
JPMSELL
Exposure
JPMAPPROVED
AMZNBUY
Daily loss
AMZNFLATTEN
Approve
Reject
Reduce
Flatten
8nsRisk check latency
1%Max single-trade risk
5%Daily loss kill switch
NoneOverride path

Backtest to live.

Two-layer pipeline: fast vectorized replay over years of tick data, then OMS-integrated simulation with realistic fills, slippage, and borrow costs before any capital goes live.

0%10%20%30%40%JanMarMayJulSepNovDec
1.82Sharpe Ratio
-8.4%Max Drawdown
63%Win Rate

LLMs research. They never touch orders.

Language models parse 10-Ks, earnings calls, filings, and social sentiment for lawsuits, management changes, and accounting red flags that break cointegration — the alpha traditional quants miss.

ACME Corp — 10-K Annual Filing
SEC EDGAR · Filed 2024-02-14
Pending litigation
CFO departure
Revenue recognition change
Extracted Risk Signals
Pending litigation
HIGH

Material risk — undisclosed class action

Confidence
92%
CFO departure
MED

C-suite turnover — 3rd in 18 months

Confidence
78%
Revenue recognition change
LOW

ASC 606 deviation — accelerated booking

Confidence
65%
Signals feed Risk Engine — LLMs never generate orders

RL agents trade. Risk engine holds veto.

Reinforcement learning agents consume market state, portfolio exposure, and microstructure features to generate trade signals — constrained by a risk engine that holds absolute veto.

Activations
Gradient Ascend
RL Rejection
Gradient Descend
memory storage
market signals

From backtest to live.

Real-time position tracking, order flow monitoring, and P&L with sub-second updates. The same OMS that runs in simulation runs in production.

dnalyaw-oms :: live-trading
LIVE· 14:32:16 EST
Unrealized P&L+$42,187.53+2.14% today
Intraday P&L
09:3012:0014:32
Open Positions126 pairs active
Fill Rate94.2%47/50 orders
Order Book · AAPL
187.401,200
187.38800
187.352,100
187.32600
187.301,500
900187.42
1,400187.45
500187.48
1,800187.50
700187.53
Spread: 0.02 (0.01%)
Recent Orders
TIMESYMSIDEQTYPRICESTATUSID
14:32:07AAPLBUY200187.42FILLED#7841
14:32:09NVDASELL150892.10FILLED#7842
14:32:11JPMBUY300214.85PARTIAL#7843
14:32:14METASELL100582.30WORKING#7844
14:32:16XOMBUY250108.67WORKING#7845
Avg Latency< 1ms
Uptime99.97%
Daily Volume$2.1M

Measure everything.

Transaction cost analysis from day one. Every fill captures arrival price vs fill price, slippage, and market impact — the gap between backtest and live.

Latency Histogram

p50p95p99

Slippage Distribution

0bpstail

Fill Quality

94.2%fill rate
p99 Latency7.8µs
Avg Slippage1.2bps
Fill Rate94.2%
Orders/Day>1M

Exchange-proximate infrastructure

Servers deployed in the same region as major exchanges for reduced latency. Direct colocation available on demand.

NY — NYSE / NASDAQ

New York region deployment near NYSE and NASDAQ matching engines. Equinix colocation available on demand.

HK — HKEX

Hong Kong region deployment near HKEX. Equities, derivatives, and Stock Connect access.

Normalized Feed

Heterogeneous data sources unified into a single schema. Quality gates filter spreads, stale quotes, and price spikes.