Commodity trading · SaaS Blockchain 3i Infotech Dubai · 2024

A blockchain-based commodity trading platform for a leading Dubai-based company.

A SaaS platform for high-volume commodity trading on blockchain — built to serve traders, market-makers, and ecosystem partners with a single auditable settlement layer. At 3i Infotech I led the end-to-end UX. I owned UX strategy, the design system, and the trader-desk experience — the highest-stakes surface in the product.

Role
UX Lead
3i Infotech
Team
1 PM · 2 designers
· 14 engineers
Duration
2024 · ~6 months
discovery → live
Outcome
Trader desk · OMS
· settlement console
DMCC Agri commodity trading platform — overview screen
Live screen Trader desk in production · platform overview
Anonymised · client UI redacted
DMCC · Trader Desk 15:47:22 GST ▲ COTTON $1,842 +0.62% ▼ COFFEE $214.50 −0.18% ▲ CARDAMOM $48.20 +1.04% CHAIN OK ORDER BOOK · COTTON-CFD spread 0.4 · depth 26 BID QTY ASK 1,841.6 12.0 1,842.0 1,841.4 22.0 1,842.4 1,841.2 8.0 1,842.6 1,841.0 26.0 1,843.0 1,840.8 14.0 1,843.2 1,840.6 10.0 1,843.6 1,840.4 18.0 1,844.0 NEW ORDER BUY SELL SYMBOL COTTON-CFD-1M QTY · CONTRACTS 100 ≈ 22.7t LIMIT $ 1,841.80 COTTON-CFD · 1H $ 1,842.00 · Δ +0.62% 1,842.0 SETTLEMENT · BLOCKCHAIN TX HASH PAIR QTY STATUS 0x84a2…f1c COTTON 100 SETTLED ✓ 0x91d8…7e2 COFFEE 50 PENDING ⋯ 0x6f1a…b04 CARDAMOM 200 SETTLED ✓ 0x2c44…a01 COTTON 75 SETTLED ✓ OPEN POSITIONS COTTON-1M +1,358 +100 @ 1,841.20 COFFEE-3M −642 −40 @ 214.50 CARDAMOM-1M +218 +200 @ 47.10 CRUDEPALM-2M −84 −10 @ 1,402.00 NET + $850 RISK · LIVE EXPOSURE $ 412k VAR · 95% $ 18.4k MARGIN UTIL 62% SETTLE QUEUE 3 Settle all eligible →
DMCC Agri · Trader desk + settlement
01 · Problem

Commodity trading is opaque, slow, and trust-starved.

Traditional commodity trading runs on bilateral phone calls, paper LCs, multi-day settlements, and trust relationships built over decades. For a Dubai-based commodity house targeting global supply chains, that model was no longer scalable. They needed a SaaS platform that brought institutional-grade trading UX to commodities — with blockchain settlement to remove the multi-day clearing and counterparty risk.

The platform's success was contingent on a second, equally hard problem: building partnerships across the commodities ecosystem. A trading platform with no counterparties is a brochure. From day one, the product had to onboard market-makers, exporters, banks, and clearing partners — each with their own KYC requirements, compliance regimes, and operational rhythms. Onboarding had to be fast enough that a partner's compliance team would say yes, and rigorous enough that the platform's compliance team would let them on.

"Our customers shouldn't have to choose between Bloomberg-grade UX and counterparty trust. We have to deliver both — a settlement layer they can audit, and a partner ecosystem they can rely on." — Client product head, kickoff

Core problems:

  • Information density without overload — traders need every depth row, but cognitive load kills accuracy.
  • Settlement must be visible — chain hash, status, and finality are part of the UX, not back-office noise.
  • Order errors are catastrophic — fat-finger protection, confirmation flows, override audit trail.
  • Multiple personas — traders, market-makers, ops, compliance — overlapping but distinct surfaces.
  • Partner onboarding is a product surface — fast for partners, rigorous for compliance, and visible to the trader who needs to know who they're dealing with.
02 · Process

Trader interviews first, then engineering reality-checks, then design.

Trading-floor UX is unforgiving. Get one trade ticket layout wrong and you've taught 100 traders to make mistakes 1,000 times a day. We started with deep contextual interviews with 8 commodity traders across two desks, and a parallel sync with the smart-contract / settlement engineering team to map what state the UI could and couldn't show.

  • Weeks 1–3 — Trader research: shadowed open-trading sessions; quantified attention patterns (eyes never leave depth + chart for > 1.5s).
  • Weeks 3–5 — Settlement reality: mapped every chain state and exception path with eng; settled on 4 user-facing settlement statuses (PENDING, CONFIRMED, SETTLED, EXCEPTION).
  • Weeks 5–8 — Trader desk design: three-pane layout, tickets in muscle-memory positions, confirmation flows tuned per order size.
  • Weeks 8–14 — System + ops: back-office consoles, compliance views, partner onboarding, reporting.
  • Weeks 14–24 — Polish, cohort tests, launch: closed beta with 12 traders → general availability for the client's institutional book.
TRADER DESK · 3-PANE LAYOUT L · DEPTH TICKET BUY SELL M · CHART M · SETTLEMENT 0x84a2…f1c · SETTLED ✓ 0x91d8…7e2 · PENDING ⋯ 0x6f1a…b04 · SETTLED ✓ 0x2c44…a01 · SETTLED ✓ R · POSITIONS · RISK COTTON +100 +1,358 COFFEE −40 −642 CARDAMOM +200 +218 RISK VaR 95% $18.4k Margin 62% Settle Q 3
Three panes · muscle memory

Layout decisions tuned to the trader's eye

Depth on the left, chart and settlement in the middle, positions and risk on the right. Tickets sit at the bottom of the depth pane — where the trader's hand already is.

Eye-tracking proxy from interviews: traders never look more than 1.5s away from depth+chart. Every secondary surface had to fit the periphery, not steal the centre.

DMCC Agri trader desk — depth, ticket, chart, and settlement panes
Live screen 3-pane trader desk · depth, chart, settlement, positions
03 · Design system

Trading UI tokens, fat-finger guards, and a settlement state model.

We borrowed institutional trading conventions where users were already trained — green = bid, red = ask, monospaced numerals, no animations on price ticks — and added a custom layer for blockchain settlement that didn't exist in our reference set.

trade.bid
#21C688
trade.ask
#E87070
surface.deep
#0A1614
surface.raised
#0E2018
chain.pending
#E8A34C
text.subtle
#5E7A72
Surfaces
5 surfaces — Trader desk, Settlement console, Ops, Compliance, Partner portal
Components
140+ trading-grade components — depth tables, tickets, blotter, chain hash chips
Fat-finger guards
Order size threshold → confirmation modal · price drift threshold → re-arm before send
Settlement model
4 user-facing states — PENDING · CONFIRMED · SETTLED · EXCEPTION · with chain hash + explorer link
Localisation & data
Multi-currency, multi-tz, RTL ready · monospaced numerals · 4-decimal precision tokens
04 · Intelligence layer

Anomaly detection, not auto-trading.

No auto-execution. AI is purely advisory — flagging weird trades, drift on settlement, and compliance risk. The trader or compliance officer always decides.

✦ Trading

Fat-finger + off-book detector

Flags orders > 3σ from the trader's 30-day median size or > X% from VWAP. Surfaces as a modal-blocking confirm with a one-line reason.

DecisionAlways trader
False-positive rate< 6%
✦ Settlement

Settlement drift watch

Predicts settlement-time outliers based on network conditions and chain mempool depth. Alerts ops before SLA breach instead of after.

DecisionOps decides
SLA breach−18%
✦ Compliance

Wash-trade pattern flag

Detects circular trade patterns across counterparties; routes to compliance with the full trail. Compliance officer is the decision-maker.

DecisionCompliance owns
Audit loggedEvery step
✦ Onboarding

Partner KYC accelerator

Pre-classifies KYC documents, extracts fields, flags inconsistencies. Cuts onboarding time from days to hours; humans verify before signing off.

DecisionCompliance approves
Time saved~70%
DMCC Agri partner onboarding and KYC console
Live screen Partner onboarding · KYC accelerator with compliance review
05 · What I owned

UX lead at 3i Infotech — trader desk & system.

UX strategy
Defined the trader-desk model, settlement state taxonomy, partner-onboarding flow, and persona-mapped surfaces across traders, ops, compliance, and counterparties
Trader desk design
End-to-end design of the highest-stakes surface — depth, tickets, chart, blotter, risk
Design system
140+ trading-grade components · fat-finger logic · chain settlement vocabulary
Engineering partnership
Daily sync with smart-contract + FE leads · co-defined the settlement state machine
Trader research
Contextual interviews with 8 commodity traders · closed-beta cohort with 12 traders
Stakeholder alignment
Weekly review with client product head · monthly steering committee
06 · Outcomes

Six months in: trader-grade UX over a blockchain settlement layer.

Illustrative numbers — I'll swap in verified figures before publishing.

5
SURFACES SHIPPED
ONE COHERENT SYSTEM
140+
TRADING-GRADE
COMPONENTS
−18%
SETTLEMENT SLA
BREACHES
−70%
PARTNER KYC
ONBOARDING TIME
Eco
PARTNERS
ONBOARDED IN LAUNCH COHORT
12
TRADERS IN
CLOSED BETA
< 6%
FAT-FINGER FLAG
FALSE-POSITIVE
4
SETTLEMENT STATES
USER-FACING
~6 mo
DISCOVERY
→ LIVE
DMCC Agri compliance and ops console
Live screen Compliance + ops console · audit trail and exception handling
07 · Reflections

What designing for real money taught me.

DO AGAIN

Co-design the state machine with engineering on day one. Settlement statuses were a UX problem disguised as a backend problem; treating them as one made every later decision easier.

DO AGAIN

Let convention do the heavy lifting. Borrowing from institutional terminals meant traders trusted the platform on day one. Novelty for novelty's sake would have killed adoption.

DO AGAIN

Fat-finger guards as a system primitive. Once we stopped treating them as a per-screen concern and made them a token-driven primitive, every order surface inherited the protection.

DO AGAIN

Treat partner onboarding as a product surface, not an admin form. Every counterparty we onboarded fast was a counterparty the traders trusted to trade against. Onboarding UX is liquidity UX — they're the same thing on different timescales.

DIFFERENTLY

Closed beta with traders earlier. We had a first cohort at week 14; week 8 would have caught two layout decisions before they got expensive.

DIFFERENTLY

Designed compliance & ops with equal weight from day one. Trader desk got 80% of attention early; ops/compliance had to play catch-up later.

LESSON

Blockchain explainability is a UX surface, not a footnote. Showing chain hash + status + finality is the difference between users believing in the settlement layer and not.