AI quant trading systems
Starter through production and flagship Apex Institutional Quant (P13–P15, P17). Engineering deliverables with risk rails, source handoff, and written acceptance — not profit, return, or win-rate guarantees.
Harbor Artificial
AI that ships in a clear scope.
We build specialty AI quant trading systems and bounded AI projects — starter through production platforms, plus RAG pilots, LLM modules, and agents — with fixed prices, clear timelines, and a written definition of done.
Built for founders and product teams who want a finished piece of work — not an open retainer.
Choose a path
Two clear ways in
Equal weight for both lanes. Pick the path that matches the outcome you need — then jump to the package anchors.
Docs, agents, sites & ops
Docs Pilot, Feature Module, Workflow Agent, FAQ bots, landings, checkout, and automations — fixed scope for SMBs and product teams who need a shippable slice.
- Fixed packages
- Written definition of done
- Deposit before build
No profit guarantees. Composite studio quotes on this site are illustrative — not attributed client testimonials.
Packages
Transparent scope, price, and timeline
Flagship and core packages we ship most: Apex Institutional Quant, AI quant trading (starter through production), Docs Pilot, Feature Module, Launch Landing, and Checkout Kit. Sticker and rush pricing. Payment terms shown up front.
Flagship
P17
Apex Institutional Quant
Most advanced Harbor quant engagement — research-grade algorithms, market data pipelines, walk-forward backtesting, portfolio construction, execution + risk, observability, and operator tooling over ~6 months. Engineering deliverable, not investment advice.
Scope, inputs & rush
Full $120,000 due before kickoff. No milestones. Source handoff after written acceptance of deliverables (already paid in full).
Rush: not offered on this engagement
In scope
- Multi-strategy / multi-factor research framework with versioned experiments
- Market data ingestion pipelines (historical + scheduled updates) with quality checks
- Feature store / signal library with reproducibility
- Walk-forward / purged CV-style backtesting harness with tear-sheet reporting
- Portfolio construction & risk (limits, exposure, drawdown controls) as coded policy
- Execution layer hooks (broker/exchange APIs or simulation) with audit logs
- Ops dashboard: PnL attribution views, risk, job health, alert hooks
- CI for research notebooks/pipelines where scoped
- Full source, architecture docs, runbooks, training sessions (bounded count)
Out of scope
- Profit / alpha / return / win-rate guarantees
- Custody of client funds
- Unlicensed investment advisory
- Unbounded live capital mandates
- 24/7 trading desk staffing
What we need from you
- Strategy thesis
- Venues
- Data sources / licenses
- Risk policy
- Success metrics for engineering acceptance (not returns)
P13
AI Quant Trading Starter
Starter AI quant trading system: automate one strategy with exchange/broker API, risk controls, optional LLM decision assist, code comments, model docs, and full source — the entry tier before multi-strategy or production platforms. Engineering deliverable, not a profit promise.
Scope, inputs & rush
100% upfront at checkout before kickoff. Scope locked to a written acceptance checklist; change orders billed separately. Blueprint sign-off authorizes build; full pay already authorizes code release at handoff.
Rush: $7,800
In scope
- One primary strategy/bot path with entry/exit logic as specified
- Exchange/broker API connectivity for the agreed venue (or simulation if scoped)
- Basic risk controls (position size, SL/TP hooks as specified)
- AI/LLM hook for decision assist as scoped (provider of your choice)
- Detailed code comments, model documentation, and full source handoff
- One revision round against the written acceptance checklist
Out of scope
- Profit, return, or win-rate guarantees
- Acting as broker, RIA, or custodian of client funds
- Knowledge graph / ontology / taxonomy systems
- Multi-strategy or production quant platforms (see AI Trading System / Production Quant Platform)
- Complex multi-cloud infra beyond agreed deployment target
- Ongoing trading ops / 24/7 monitoring retainer
What we need from you
- Strategy goals, markets/symbols/timeframes, and entry/exit rules
- Primary exchange/broker (or simulation-only)
- Risk limits and deployment preference
- Any existing code, Pine/TradingView, or specs
P14
AI Trading System
AI trading system with multi-strategy logic, centralized risk, backtesting, exchange execution hooks, documentation, and source — built to your rules, without financial-return guarantees.
Scope, inputs & rush
100% upfront at checkout before kickoff. Scope locked to a written acceptance checklist; change orders billed separately. Blueprint sign-off authorizes build; full pay already authorizes code release at handoff.
Rush: $15,000
In scope
- Multi-strategy / multi-bot logic as bounded in kickoff (count agreed in writing)
- AI/LLM decision components with chosen provider/model
- Risk management layer and order/position handling
- Historical backtesting path for agreed symbols/timeframes
- Exchange API integration(s) in scope
- Monitoring/alerts basics as scoped
- Code comments, model docs, source; two revision rounds
Out of scope
- Profit, return, or win-rate guarantees
- Acting as broker, RIA, or custodian of client funds
- Knowledge graph / ontology / taxonomy systems
- Full production multi-agent platform with ops dashboard suite (see Production Quant Platform)
- Unbounded strategy count
- 24/7 retained trading ops
What we need from you
- Strategy descriptions and agreed strategy/bot count
- Exchanges and markets/timeframes
- AI provider/model preferences
- Risk rules and validation data/examples
- Deployment target
P15
Production Quant Platform
Production-minded quant platform with AI agents, automated execution, risk controls, backtesting, and an ops dashboard — engineered for auditability and handoff, not promised alpha.
Scope, inputs & rush
100% upfront at checkout before kickoff. Scope locked to a written acceptance checklist; change orders billed separately. Blueprint sign-off authorizes build; full pay already authorizes code release at handoff.
Rush: $23,400
In scope
- Production architecture for agreed strategies/bots and markets
- AI agent components for decision/workflow as scoped
- Automated execution + position/order management
- Centralized risk controls and validation/backtesting lab path
- Ops dashboard (performance, bots, status) as scoped
- Hardening: logging, config, security basics for agreed deploy target
- Docs, comments, source; three revision rounds
Out of scope
- Profit or return guarantees
- Regulated finance system-of-record / custody
- Knowledge graph / ontology / taxonomy systems
- Open-ended R&D
- Multi-month retained trading desk
- HIPAA / clinical work
What we need from you
- Full goals and feature priorities
- Strategies, exchanges, and markets
- AI stack preferences
- Risk and execution rules
- Deployment environment
- Validation datasets
- Interface and workflow needs
P1
Docs Pilot (RAG)
A chat-with-your-docs pilot — retrieval, citations, and a written checklist for what “good” looks like.
Start payment Pay balance
Mid/final links are backup. Harbor issues the invoice + pay link automatically at each gate (mid-review / finished). Final pay authorizes source/code release — not a silent card charge.
Scope, inputs & rush
50% deposit at checkout to start; 50% final when finished — before source/code handoff. Blueprint sign-off authorizes build; final pay authorizes code release.
Rush: $4,800
In scope
- Corpus ingest pipeline for a bounded document set
- Retrieval + grounded answer path with citation links
- Simple web or API surface for pilot users
- Acceptance tests for retrieval quality and refusal behavior
- Handoff runbook (env, rebuild, known limits)
Out of scope
- Enterprise SSO / multi-tenant auth
- Guaranteed zero-hallucination answers
- Ongoing corpus ops or 24/7 support
- HIPAA / clinical or regulated finance systems of record
What we need from you
- Your documents (or a sample) and how we can access them
- Who it’s for, plus 5–10 example questions
- Hosting preference (your cloud or local)
P2
Feature Module (FastAPI + LLM)
A FastAPI feature module that wraps an LLM behind clear request/response contracts — ready for your team to own.
Start payment Pay balance
Mid/final links are backup. Harbor issues the invoice + pay link automatically at each gate (mid-review / finished). Final pay authorizes source/code release — not a silent card charge.
Scope, inputs & rush
50% deposit at checkout to start; 50% final when finished — before source/code handoff. Blueprint sign-off authorizes build; final pay authorizes code release.
Rush: $6,000
In scope
- FastAPI service or module with typed request/response schemas
- Prompt + tool or structured-output path as scoped
- Logging, retries, and failure modes documented
- Acceptance tests against agreed scenarios
- Handoff: OpenAPI surface + runbook
Out of scope
- Full product UI redesign
- Realtime voice / streaming telephony
- Founding-engineer retainer or on-call
- Regulated finance system-of-record writes
What we need from you
- A draft API contract or user story
- Model / provider preference if any
- Existing auth or gateway constraints
P6
Launch Landing
One high-clarity landing page with lead capture and basic SEO — ready to share in about a week.
Start payment Pay balance
Mid/final links are backup. Harbor issues the invoice + pay link automatically at each gate (mid-review / finished). Final pay authorizes source/code release — not a silent card charge.
Scope, inputs & rush
50% deposit at checkout to start; 50% final when finished — before source/code handoff. Blueprint sign-off authorizes build; final pay authorizes code release.
Rush: $1,800
In scope
- A single page: hero → proof → CTA
- Form or email capture
- Basic SEO (title, meta, Open Graph)
- Deploy to your agreed host
Out of scope
- Multi-page marketing site (see Marketing Site)
- Paid ads management
- Deep CRM automation (see Ops Automation / API Glue)
What we need from you
- Offer one-liner and who it’s for
- Where leads should go
- Brand assets (or defaults OK)
P10
Checkout Kit
Stripe Checkout or Payment Links, success and cancel pages, a webhook stub, and a customer portal link — you stay the merchant of record.
Start payment Pay balance
Mid/final links are backup. Harbor issues the invoice + pay link automatically at each gate (mid-review / finished). Final pay authorizes source/code release — not a silent card charge.
Scope, inputs & rush
50% deposit at checkout to start; 50% final when finished — before source/code handoff. Blueprint sign-off authorizes build; final pay authorizes code release.
Rush: $2,160
In scope
- One checkout path (Checkout or Payment Links)
- Success and cancel pages
- Webhook stub that acknowledges and logs
- Customer Portal link
- Test → live checklist and runbook
Out of scope
- Custom card forms or storing card data
- Full subscription billing product or tax engine beyond Stripe defaults
- Marketplace payment splits
- Harbor as processor or fund holder
What we need from you
- Your Stripe account (you’re the merchant of record)
- Product / price IDs or a create-in-test plan
- Success and cancel URL rules
Plus more AI, web, ops, and checkout packages — see all 16 → · Custom Bundle (20% off; 5% with P17) · Custom Project (fixed quote) · Guides
Case studies
Case Studies
Fixed-scope AI and quant systems we ship: trading consoles, RAG pilots, agents, landings, and ops workflows. Each study maps to a package with timeline, scope, and handoff.
Case Study Quant
Meridian Apex
Flagship institutional quant engagement for research and trading teams that need a production-minded research-to-ops stack they can own.
- 26 wk build
- $120k fixed
- Source + runbooks
Case Study Quant
Signal Deck Starter
Single-strategy AI quant starter for teams that want broker hooks, auditable risk rails, and an LLM assist lane they can operate from day one.
- 1.5–2 week build
- Fixed scope
- Full source handoff
Case Study Quant
Harbor Lattice
Multi-strategy AI trading system for teams that have outgrown a single bot and need shared risk plus a historical validation path.
- 3 week build
- Shared risk layer
- Backtest path
Case Study AI
Ledgerly Docs Pilot
Chat-with-your-docs RAG pilot for teams that need grounded answers with citations — not a freeform chatbot.
- 2 week build
- Citations required
- Written acceptance
Case Study AI
Relay Feature Module
FastAPI + LLM service module with typed contracts and documented failure modes for product teams that need something engineering can own.
- 2 week build
- OpenAPI handoff
- Typed contracts
Case Study Agents
Conduit Workflow Agent
Bounded workflow agent for teams that need allowlisted tools, step limits, traces, and pass/fail scenarios agreed before build.
- 3 week build
- 40/30/30 milestones
- Trace observability
How we work
Discover · Build · Review
A simple path: align on “done,” build to that list, then accept with a short bugfix. Deposit before we build; change orders in writing.
-
Discover
We align on the outcome, what you’ll provide (docs, APIs, access), constraints, and a short written “done means…” checklist — before any build deposit.
-
Build
We build to that checklist in a fixed window. Longer packages include a mid-build check-in. Anything outside scope gets a written change order with price and timing before extra work starts.
-
Review
You accept against the agreed tests (pass/fail). You get a handoff runbook, plus a 14-day bugfix on what we delivered.
Trust
What working with us feels like
Fixed packages. Written definition of done. Deposit before we build. These launch composites show the kinds of outcomes Harbor is built for. Illustrative voices, not named case studies.
“Honestly we just needed our people to ask questions against our own docs and see where the answer came from. Harbor did that in two weeks for a fixed price. Checklist said done and it was done. No drama.”
“I'm not technical. I am the person stuck in Slack at 4pm because a lead never made it into the CRM. They set up two workflows, error alerts, and a runbook so my team can rerun stuff without texting me. Total lifesaver.”
“Solo. Money is tight. Checkout Kit got Stripe live in a week and I stayed the merchant of record. That mattered more than a pretty page.”
“Eval harness in a week. We stopped shipping prompt tweaks on vibes. Finally.”
“Five pages, real SEO basics, one clear email CTA. No CMS rabbit hole. We launched on the timeline that was on the invoice, which almost never happens with agencies.”
“Offer page Monday. Stripe path by Friday. Launch Stack was the first time shipping the offer did not mean three freelancers and a spreadsheet fight.”
Launch composites for a new studio. Illustrative voices only. Not attributed to named clients.
Written acceptance
Every package ships against a short done means checklist you agree up front. Acceptance is pass/fail on that list, not open-ended polish. You also get a handoff runbook.
Deposits before we build
Work starts after the deposit clears. Most packages are 50/50: final pay when finished, before source handoff. Longer builds like Workflow Agent use a mid-review invoice. Quant packages (P13 through P15 and flagship P17) are 100% upfront at checkout. Blueprint sign-off authorizes build; final pay authorizes code release.
What we don’t take on
Better to know early. We pass on HIPAA or clinical work, regulated finance systems of record, realtime voice products, and open-ended founding-engineer retainers. If that is your need, we will say so plainly and kindly.
Questions before you pick a package?
Write us at harborartificial@pm.me. Share what you’re trying to ship, your timing, and any constraints you already know.
Email usNot a fit?
Here’s what we don’t take on
Better to know early. If your need is on this list, we’re probably not the right partner — and that’s okay.
- HIPAA or clinical work
- Regulated finance systems of record
- Realtime voice products
- Open-ended founding-engineer retainers
Want to see if a package fits?
Share what you’re trying to ship, your timing, and any constraints you already know. We’ll reply from harborartificial@pm.me.