Financial Services

Modernizing a Live Derivatives Trading System at Citi

Citi needed faster pricing models in a live trading environment where downtime or calculation errors could halt trading. We modernized the platform incrementally, delivering real-time accuracy without disrupting active trades.

Services Provided

Web Development, Data Engineering, Application Modernization

Product Type


API-based Web Platform

Technologies Used


Kotlin, React, Typescript

Project Highlights

Traders always access the latest approved pricing models — no stale data risk

Platform now hosts 9 tools used by 6 teams across the organization

Zero disruption to trading operations during rollout

About

Citi's mission is to serve as a trusted partner to their clients by responsibly providing financial services that enable growth and economic progress. In derivatives trading, pricing accuracy is everything — a stale model or calculation error doesn't just cause inconvenience, it can mean significant financial losses or regulatory exposure.


Challenge

Traders faced significant challenges accessing the latest derivative pricing models. The models lived on local machines, requiring manual downloads and Python knowledge to use. This created multiple risks: traders might use outdated model versions, downloads took time during fast-moving markets, and the decentralized approach made it impossible to ensure everyone was using approved models.

The real danger wasn't just speed — it was consistency and accuracy. In derivatives, using a stale pricing model or an unapproved version could mean mispriced trades, regulatory findings, or losses that compound quickly.



Solution

Def Method built a centralized web platform that hosts the latest Quant libraries and integrates vol-swap and collateral-based pricing support. The platform is API-based, so traders access models through the web without downloading anything to their machines.

This architecture solves the consistency problem: every trader automatically gets the latest approved version. No stale data, no version conflicts, no need for Python knowledge. The team designed for minimal disruption, rolling out incrementally so trading workflows were never interrupted.


Results

The platform eliminated the stale-model risk entirely — traders now always access the latest approved pricing models instantly. No more downloads, no more version mismatches, no more Python requirement.

What started as an experiment by a small engineering team has grown into a platform hosting 9 different tools used by 6 teams across the organization. The platform is now fully supported with a team of 10 engineers working to incorporate additional tools and models.

Most importantly: zero disruption to trading operations during the entire rollout. Traders got faster, safer access without any interruption to their workflows.