My personal blog
This is my personal website. Content on here spans my personal, professional, and philanthropic interests. Personal interest postings will be drawn from current events, readings, book lists and other activities that influence my personal viewpoints. Separate from my personal viewpoints, I will post DevStories which are short musings of work I have done in the…
DevStory #17: Performant low code streamlit based full text search unlocking data from 900 million pages / 150 million Texas land records
Streamlit based full text search application to facilitate deep search on 150 million TIFF images containing information about natural resources in rural Texas. An easy way to extract and search this information using custom AI tools was developed as part of this prototype. The generated text corpus results in options to build a significant number…
DevStory #16: Image Enhancement, OCR / ICR , NLP, Full text search on information goldmine extracted from 900 million pages / 150 million documents
An information goldmine exists in about 150 million TIFF images containing information about natural resources in rural Texas. An easy way to extract and search this information using AI was developed as part of this prototype. The corpus generated results in options to build a significant number of value-added applications that benefit the real estate,…
DEVSTORY #11 : Counterparty Knowledge Graph (CKG) for up to date information on your trading counterparties
Before you know it counterparty information and other reference data is out of date. As the data ages, a larger proportion of the data gets out of date. The CKG is a collaborative robotic process to keep all ETRM systems in sync using a single Knowledge graph updated by each counterparty in a trading network. Robotic process automation then keeps all ETRMs in sync using event driven automation
DEVSTORY #13: Pricing CME Natural Gas American Options (ON) using options pricing functions as a service library
CME Natural Gas American Options (ON) priced using an American Call Option (Bjerksund Stensland 2002 approximation) model for live pricing and greeks. Low code, HA, hybrid cloud and kubernetes deployed containerized model accessed from any where through a https endpoint in a performant manner
DEVSTORY #12: Using function as a service (faas) to retrieve data from a knowledge graph
A trading firm stores its reference information in the Counterparty Knowledge Graph (CKG). Settlement contacts are one category of information of interest and queryable from the Counterparty Knowledge Graph (CKG). To expose the list of contacts related to the settlements role, for a particular commodity at the trading firm, there is a need to provide an endpoint that is manually or programmatically queryable.
DEVSTORY #8: Greenfield natural gas portfolio mid-office profit & loss platform for startup natural gas trading desk
mid office required a system to capture trades, capture trader marks and report daily p&l to match trader estimates. An operational, reliable, cheap system was required for this USD 30 million dollar startup portfolio. The long term plan was to migrate to a traditional ETRM system as the portfolio grew to USD 120 million. Since the startup decided to move quickly, the time and cost of evaluating and implementing a traditional ETRM would set then back several months if not a year at best.
A quick comprehensive and complete solution implemented in 2 months was delivered that met needs of desk.
DEVSTORY #14: Quantitative analysts rollout derivative pricing models globally to desktop pricers, ETRM, Accounting, and Risk systems in minutes not months
Roll out of enhanced pricing algorithms to price tail risk better, took three months to be deployed globally at a large trading firm. In the interim period, market makers, risk managers, and quants were using the new model. The mid-office was reporting off a nonvalidated model provided by the ETRM vendor. The back-office was generating…
DEVSTORY #15: RDBMS data cleansing: Using pandas/modin instead of SQL to clean ISO data for FTR desk
Clean data makes data scientists more efficient. Using simple tools data wrangling tools like Pandas/Modin, DASK and Vaex instead of traditional SQL procedures can result in cleaner data and better analytics results.
DEVSTORY #14: Quantitative analysts rollout derivative pricing models globally to desktop pricers, ETRM, Accounting, and Risk systems in minutes not months
A low code, highly available, planet-scale, containerized, Kubernetes deployed, functions as a service approach, that allowed the quant and IT teams to roll out models using a Jenkins-based CI/CD process several times a day. Models were deployed as stateless functions called using simple HTTP get requests accessible from software clients such as spreadsheets, javascript, python, Java, C++, C#, and Golang programs. Several research quant groups could leverage these models from tools like R, SAS, and Matlab.
DEVSTORY #11 : Counterparty Knowledge Graph (CKG)
Schedulers, Settlement analysts, and AR/AP professionals at a commodity trading operation need up-to-date operational reference data related to counterparties and service providers. Most systems of record used today by personas in these operational roles are out of date as soon as they are deployed as data changes quickly and the systems are reliant on people to keep them up to date.
Keeping reference data up to date across multiple firms calls for a central repository of this information. The information is updated once by the responsible party. Notifications of change can be sent to other business entities that can elect to update their systems of record manually or programmatically on some or all notifications.