IgniteEdge Field Guide
Playground Fund program
Edition 01 · Capital markets, run locally11 entries · one platform

A field guide to you.

Find your role in capital markets. Each entry shows what IgniteEdge does for you, with a worked example you can hold up against your own desk.

Every example runs on a computer you already control: your workstation, your firm's server or your GPU cluster. There is no vendor cloud in the middle.

The local computer story

Same software. Your machine, at any size.

Start on one desk and grow to a cluster without changing where your data lives. Nothing is sent to a vendor cloud at any step.

Workstation one trader · one GPU Firm server an RIA or small desk GPU cluster broker-dealer · fund vendor cloud
01
Fund management company · formation

Building the relationships to launch a fund program

Allocators, a prime broker, administrators and counsel need to take a systematic program seriously before a dollar is raised.

We lead with a working system: live services and a documented operating model. Launch materials are shared only with qualified investors, under NDA.

Live service endpointsOperating modelInvestor materials (NDA)In progress
CounselAdministrator Prime brokerAllocatorsFirst close each meeting opens with the live platform, not a deck
The path we are walking nowIllustrative
AGENT PROPOSAL · SAMPLE Buy index put spread, 10 lots Margin check (Mercurius)PASS Semantic check (local LLM)PASS Signed to ledger (HMAC)OK Human approvernamed PM APPROVED then 14-state OMS lifecycle
Nothing routes without the stampIllustrative
02
Fund sponsor · after approvals are in place

Raising capital for agentic trading with a human in the loop

Investors want to know who is in control. Agents propose; a named person approves.

Every proposal is margin-checked, written to an HMAC-signed ledger and routed through a 14-state order lifecycle, so each decision can be audited.

SparkMercuriusAgentic OMSHITL queuePaper pilot running
03
Research desk · self-trained models

Self-trained agentic trading on owned hardware

A cluster of NVIDIA Grace-Blackwell nodes on a 200 Gb/s RDMA fabric serves open-weight models through vLLM and TensorRT-LLM.

Fine-tuning, fast and slow reasoning, and model routing all run locally. Strategy IP never leaves the building.

DGX Spark clustervLLM · TensorRT-LLMNVFP4 / FP8Fine-tuningLive
Node A · fast pathNode B · slow path quick checksdeep reasoning 200 Gb/s RDMA your racks · your training data
Two paths, one buildingIllustrative
Margin lending partner

Margin intelligence and portfolio ideas for your clients

RegT, Portfolio Margin and SPAN 2 requirements are computed before the trade, not after. Spark's recommendations already respect your house limits.

In the example, a client's futures book is priced with the four-part SPAN 2 formula before the order leaves. The requirement fits the available collateral, so the idea goes through.

Mercurius firewallSPAN 2UnifiedRBMPortfolio recsLive APIs
04
PRE-TRADE REQUIREMENT · SAMPLE BOOK x · HVaR 0.7 × 150,000105,000 (1−x) · Stress 0.3 × 210,00063,000 Liquidity add-on11,400 Concentration add-on4,800 Requirement$184,200 Available collateral$250,000 Within limit · 74% used · order may proceed
Checked before it tradesIllustrative numbers
05
Automated portfolio construction

Unattended long/short, index hedges and 0DTE income

ONE SESSION, THREE SLEEVES L/S book Index hedge 0DTE income 09:58 batch15:45 flatten each sleeve: own limits + kill switch
A day on the scheduleIllustrative

A nightly-researched long/short equity book is executed as one batch and flattened by the close. An index-put hedge sits underneath. A 0DTE SPX sleeve runs on top for income.

Each sleeve has its own limits and kill switch. Every order crosses the same margin firewall.

L/S equityIndex-put hedge0DTE SPXieBalancer · tax-awareYour broker-dealerDemo / paper accounts
06
Broker-dealer · sovereign AI

Own your AI stack for confidential portfolio calibration

Models, the RAG knowledge base and the analytics install inside your data center. Volatility surfaces, SABR smiles, regime detection and PnL attribution run on your data, under your controls.

On-prem LLMsRAGSABR · term structurePnL attributionDeployable
YOUR FIREWALL Your book + data CalibrationSABR · regimes Insights
No step leaves the dashed lineIllustrative
WHAT YOU HOLDWHAT THE AGENT SEES Client Aacct ••••4821stays in your CRM tok_7f3a91positions · riskno name, no acct Allowed: research · explain · propose Blocked: identity · account numbers permissions set per client by the firm
Swapped at the gatewayIllustrative
07
Broker-dealer · client experience

Agentic portfolios for your clients, without their identity

Keep your platform and your brand. Clients reach our agents through a privacy gateway that swaps names and account numbers for anonymous tokens.

Agents see positions and risk, never the person. Read-only research modes plus MCP and llms.txt discovery let you set exactly what each client can do.

Privacy gatewayAgent web clientMCP · llms.txtWhite-labelWorking
Any broker-dealer · evaluation

Test us: a fast strategy build on your market

Bring a market and a goal. We stand up a candidate strategy quickly: equity long/short, index options, CBOE binary options (XSPBW / QSB), or 0DTE and binary structures for cash flow.

You see the research, the margin impact and the order path end to end on paper, before capital moves.

Binary Options pricerSSETFStrategy sprintPaper replayAvailable now
08
Your brief Candidate Margin impact Paper run Go / no-go
Evidence before capitalIllustrative
09
Registered investment advisors

Margin-aware model portfolios and client proposals

An advisor runs model portfolios across a book of clients. IgniteEdge proposes rebalances that already respect each account's margin and risk limits, then drafts the client proposal in plain language.

Client names and records stay in your own CRM and network. The models only see anonymous tokens.

Rebalances run through ieBalancer, which applies the tax rules you set with each client's tax advisor: loss harvesting thresholds, short-term gain protection, realized-gain caps and household wash-sale checks. It applies rules; it does not give tax advice.

In the example, a 60-client book is scanned overnight. Nine accounts drift outside their model, and three of those would breach a margin limit if rebalanced the obvious way, so they get an alternative trade instead.

Model portfoliosieBalancer · tax-awareMargin firewallProposal draftsRuns on a firm serverDesign partners wanted
60 accounts · 9 drifted · 3 re-routed for marginIllustrative
ORDER TICKET · PRACTICE EXAMPLE SPX 0DTE put credit spread Sell 5850 put / Buy 5845 put× 2 Net credit1.10 ($220) Width × 100 × 2 − credit$1,000 − $220 Requirement$780 Margin check · pass Local model review · pass Your approval · waiting Approve
Max loss equals the requirement: $780Illustrative numbers
10
Independent traders

Your strategies, your model, your account

Run strategies and a local model on your own workstation or a single GPU box, connected to your own broker-dealer account. Agents prepare the ticket and check the margin. You press approve before any order is sent.

Workstation or one GPUYour broker accountApproval before ordersDesign partners wanted
11
Independent financial services professionals

A plain-language research desk for planners, family offices, CPAs and insurance pros

Ask questions in everyday words and get structured, source-labeled research cards back. The desk is read-only: it cannot place orders or touch client accounts.

Read-only agent clientSource-labeled answersNo ordersDesign partners wanted
How exposed is a retiree's 60/40 portfolio to a 2-point rise in rates? Estimated hit: about −5% overall Duration ~6 on the 40% bond sleeve → ~−12% on bonds, ~−4.8% overall Sources: holdings you supplied · public yield data · duration model READ-ONLY
Answers, not ordersIllustrative
Capability map

One platform under all eleven

Component0102030405060708091011
SparkIdeas, insights, RAG
MercuriusPre-trade margin firewall
Agentic OMS14-state lifecycle, signed ledger
SPAN 2 · UnifiedRBMFutures and cross-asset margin
Local GPUsInference & training on your hardware
ieBalancerTax-aware multi-account rebalancing
Binary Options · SSETFXSPBW/QSB, single-stock ETFs
Privacy & agent protocolsTokenized PII, MCP, llms.txt
Read-only agent clientPlain-language research desk
Investor materialsUnder NDA, qualified investors
Found yourself?

Let's talk.

Forty-five minutes. We walk the platform live, you name your market and constraints, and we leave you with a scoped pilot. Nothing is installed and no client data is shared on the call.

[email protected]
  1. Discovery call. Your platform, markets, risk policy, and where agents help.
  2. Strategy sprint. A candidate L/S, index-option, 0DTE or binary strategy on your market.
  3. Paper pilot on your computer. Workstation, firm server or cluster, with approval on every order.
  4. Go / no-go with evidence. Margin impact, fills and decision logs you can audit.

Examples marked illustrative use made-up numbers to show how the platform works; they are not real accounts, trades or results. Stages describe the state of our own systems; paper and demo results are not indicative of future performance. Any fund program is subject to the required registrations and approvals. Nothing here is an offer to sell or a solicitation to buy any security, and nothing here is tax advice. Third-party names are used only to describe compatibility.