Concept · not shipped · speculative prototype Institutional fintech · hedge-fund PM cognition Production-grounded · original IP

TradeX · Hedge Fund

A terminal designed around the manager's cognition — not a wall of widgets.

What would a hedge-fund terminal look like if you rebuilt it around the three cognitive modes a PM moves through in a trading day — rapid scan, deep investigation, execution decision — instead of inheriting the everything-at-once data-density premise? Grounded in FIX 4.4 production work at ACY Connect, shaped by conversations with the people who live the problem.

11
Canonical screens designed
3
PM cognitive states · Scan → Investigate → Execute
4-tier
AI autonomy · observe / propose, human decides
5
Screen KYC + AML flow
$1T+
AUM segment addressed
TradeX · Fund Performance & Risk Analytics
TradeX Hedge Fund terminal — portfolio risk matrix, factor decomposition, and macro correlation panels
Concept · all data simulated

The contribution · at a glance

Product designer — a terminal built on PM cognition.

A concept trading platform designed around the portfolio manager's mind — the three mental states a PM moves through, not a wall of widgets. Figures describe design scope; the practitioner research is informal and small-n; no shipped outcomes are claimed. Every number carries its status.

11Canonical screens designed — Live Markets · Portfolio · Analytics · News · Trading Desk · Compliance · Performance · Institutional Dashboard · Screener · Factor Analysis · Factor Exposure.Concept · design scope
3PM cognitive states modelled — Scan → Investigate → Execute — the spine the whole terminal adapts to.Design hypothesis
4Unmet practitioner demands — causation over correlation — sourced from practitioner research.Directional
5Screen KYC + AML verification flow — the compliance spine an institutional onboarding needs.Concept · compliance scope
4-tierAI autonomy framework — Ambient → Co-pilot → Assertive → Autonomous. A governance switch, not a smartness slider.Agents observe · human decides
n=3Informal PM conversations (L/S equity funds, $200M–$2B AUM) — plus peer review by two CFA charterholders.Directional, not validated

The paradox

More data than any trader in history — and still blind to what matters now.

A PM in 2026 has institutional-grade terminals, a proprietary quant stack and six monitors. Yet when a 2am Taiwanese-semiconductor announcement starts rippling through their carry-trade positions, no terminal shows them the causal chain — only the price move, after the damage is done. The market has reached a strange equilibrium: everything is on-screen; nothing tells you what matters right now.

The incumbent terminal gives you the data. The synthesis, the ranking, the “what to ignore during a crisis” is left to the user. TradeX is a concept exploration of breaking that equilibrium — not with more features, but with a fundamentally different architectural premise: the terminal adapts to the PM's cognitive state, not the other way around.

Live · the three cognitive states

One terminal, three minds.

The same PM, the same book — but a morning moves through three distinct phases, and each one needs a different surface. Switch the state below and watch the terminal reconfigure: Scan strips to what changed overnight; Investigate traces the causal chain and decomposes the factor exposure; Execute collapses analysis and action into one ticket with the pre-trade risk inline.

TradeX · PM cognition
Scan — 40+ positions and the whole macro in one glance. Anomalies pre-attentive; detail on demand.

All data simulated. The point is the information architecture — how the surface changes shape to match what the PM is doing — not a live feed.

Research foundation · what I actually studied

Concept informed by research — honestly hedged.

This started from real conversations, not desk research. The evidence base is documented — and its limits are named. Small-n and directional, not a formal study; but grounded in production and reviewed by people who hold the qualification.

Source 01 · Production

ACY Connect FIX 4.4

Direct exposure to institutional order states (ExecutionReport 35=8, Snapshot 35=W) and prime-broker integration on ACY's live institutional product. Grounds the execution-flow design.

Source 02 · Practitioners

3 PM conversations

Informal, directional talks (n=3, not structured interviews) with a PM, a quant analyst and a buy-side risk officer at L/S equity funds ($200M–$2B AUM), 2024. Shaped the 3 cognitive states and 4 unmet-demand categories.

Source 03 · Competitive

Incumbent terminals

Extended analysis of publicly documented IA, keyboard-command systems and panel layout — via developer docs and trial access — before any screen was drawn. Grounds density and keyboard-first principles.

Source 04 · Peer review

Two CFA charterholders

The risk-model framing and PM-workflow assumptions were reviewed with two CFA charterholders (Level III passed). A directional endorsement of the thesis — not a formal validation, and not investment advice.

Source 05 · Demand signal

~7 feature requests logged

Seven practitioner-sourced feature requests have accumulated against the prototype. Small-n and directional — a signal that the problem is real, not a validated backlog. I label it as exactly that.

Source 06 · Published

Literature & prior art

CFA Institute on decision-making under uncertainty; Kahneman (2011) + Lo & Repin (2002) on crisis cognition; TauricResearch TradingAgents (arXiv:2412.20138); Fincept Terminal v4 (37 production AI agents) confirming committee architecture is deployable.

Honest caveat: the “4 failure modes” and “3 structural barriers” are documented patterns from these sources, not primary-research findings. I'm not a quant — I understand the problem from the people who live it.

Design thesis · three decisions

The terminal adapts to the manager — not the reverse.

Decision 01

Conclusions before evidence

Surface what the system concluded first; present the supporting data second. Managers should challenge synthesis — not perform it by hand under time pressure.

Decision 02

Explainability over efficiency

Every AI signal shows its reasoning chain. At no AUM level is “the model said so” an acceptable answer — least of all at the execution boundary.

Decision 03

Decision and action in one space

Context-switching between analysis and execution is where alpha leaks. Every view connects directly to the execution pathway it naturally leads to.

What managers actually want

Intelligence that maps causation, not just correlation.

Four gaps came up consistently across the conversations that shaped this project. Current platforms give managers data; what they want is the causal chain — before the price move, not after.

01
Demand 01

Causal chains, pre-move

“Show me what's about to move my book and why” — the driver graph, not the post-hoc price.

02
Demand 02

Synthesis, not raw feeds

Rank what matters right now. The “what to ignore during a crisis” is the hardest and most valuable call.

03
Demand 03

Explainable factor risk

Factor and concentration exposure with the reasoning attached — traceable to a model, defensible under review.

04
Demand 04

Analysis → action, unbroken

The path from insight to order without a tab switch — because the switch is where conviction and alpha decay.

The hardest question · AI proactivity

How much should the AI say unprompted?

The single hardest design decision: how much the AI surfaces on its own versus answers on request. Get it wrong and you either miss the 2am signal or you cry wolf until the PM mutes it. The answer is an explicit four-tier autonomy framework — a governance switch the desk sets per signal class, not a smartness slider. Move the tier; watch the same signal change how loudly it speaks.

Autonomy tier · desk-configurable
01
Ambient
Background indicator. No interruption — a faint mark the PM can choose to read.
02
Co-pilot
Surfaces a relevant card when context fits. Present, not pushy.
03
Assertive
Interrupts with a high-confidence alert. Reserved for signals the desk pre-approved to break focus.
04
Autonomous
The system's reasoning is what the PM sees first; the raw data sits second. Still the PM's decision.
AI is a governance switch, not a smartness slider. The model observes and proposes; the PM decides. The tier controls how much focus a signal is allowed to break — a compliance and trust decision, not a capability one.

The 11 screens · failure → decision

Each screen answers a documented failure in today's tools.

Eleven canonical screens, each paired with a failure mode in current institutional terminals and the design decision that answers it. A selection of the real prototype surfaces below — click to enlarge.

TradeX · Live Markets
TradeX Live Markets — indices, watchlist, and a movers table for the morning scan
01 · Live Markets
The morning scan surface
Failure: every feed at equal weight — no sense of what changed overnight.Decision: movers + watchlist ranked so the anomaly is pre-attentive.
TradeX · Portfolio Risk Matrix
TradeX Portfolio Risk Matrix — Sharpe, VaR, Alpha, Beta across funds with sector, stress and regional panels
07 · Fund Performance & Risk
Cross-fund risk, one page
Failure: risk lives one tab away from the P&L that triggered the check.Decision: Sharpe / VaR / Alpha / Beta + stress + regional exposure on one surface.
TradeX · Factor Exposure
TradeX Factor Exposure & Attribution — style-factor Z-scores, t-stats, sector active weights, macro factor exposures
11 · Factor Exposure & Concentration
Explainable factor risk
Failure: a factor number with no reasoning is undefendable under review.Decision: Z-score, t-stat, attribution and risk contribution per factor — traceable.
TradeX · Trading Desk
TradeX Trading Desk — execution surface connecting analysis to order entry
05 · Trading Desk
Analysis → action, unbroken
Failure: conviction decays in the tab-switch from insight to order.Decision: every view routes straight to the execution pathway it implies.
01–04Live Markets · Portfolio · Analytics · News — the scan-and-context layer
05–07Trading Desk · Compliance Screener · Fund Performance & Risk — the act-and-govern layer
08–11Institutional Dashboard · Screener · Factor Analysis · Factor Exposure & Concentration — the decompose-and-defend layer

KYC & compliance

Institutional onboarding is a five-screen spine.

A hedge-fund terminal that can't onboard a fund under KYC/AML is a demo, not a product. The flow carries entity verification, beneficial-ownership, source-of-funds and suitability — informed by (not audited against) MiFID II Art. 27 and SEC 17a-4 retention.

Referenced from the production compliance pattern library at ACY Securities; the hedge-fund-PM workflow itself is the concept extension.

TradeX · Compliance Screener
TradeX Compliance Screener — KYC / AML verification surface

Status & honesty

What's designed — and what is deliberately not built.

Concept design exploration · not in production · speculative prototype. Built from production-adjacent work — ACY Connect FIX 4.4, public incumbent-terminal architecture analysis, and informal practitioner conversations. Original IP, no client overlap. Read as a research artefact.

◆ What I designed

0111 canonical screens across scan, act and decompose layers.
023 cognitive states framework — Scan → Investigate → Execute.
034 unmet-demand categories sourced from practitioner research.
045-screen KYC + compliance flow and a 4-tier AI autonomy framework.
05Visual language — density tiers, semantic colour, keyboard-first interaction.

○ Deliberately not built

—Live market data — every number is simulated.
—Real FIX execution path — referenced from ACY Connect production, not rebuilt.
—Validated quant models — HHI / VaR / regime are visualisation primitives, not calibrated.
—User testing with named PMs — directional conversations only (n=3).
—Regulatory certification — informed by, not audited against, MiFID II Art. 27 + SEC 17a-4.

Everything is on-screen. Nothing tells you what matters right now. That gap is the whole design.

Ed Chen · Senior Product Designer · a terminal that adapts to the manager's cognition, not the reverse

Explore further

The dream platform — concept, prototype, and a real practitioner pull.

TradeX Institutional (execution layer) → ACY Connect · FIX 4.4 (production) Nova · AI trust → 90-second recruiter scan
Portfolio threads

Where this case study sits in the larger web

Every problem we solve for clients has multiple valid approaches — different costs, different ROI, different risk profiles. These threads show how the approach on this page compares to others in the portfolio.

Thread

Concentration, Risk & Agents

Portfolio-level math primitives — HHI, beta, VaR, regime — rendered into UI defaults and AI-assisted decision surfaces.

Thread

Regulatory Routing & Disclosure

How upstream regulation and macro prints become downstream product defaults and Legal-safe disclosure.