https://gaultier.github.io/blog/body_of_work.html

ME

profile

skills

INFRA

AWS: Lambda, CloudWatch, EC2, IAM
PaaS: Cloud Foundry, Platform.sh
IaC: Terraform, Ansible
operating systems: Linux
servers: (Nginx, Gunicorn)
Algolia

LANGUAGES

Python: Django/DRF, Flask, FastAPI + libraries (bpython, Black, bandit, coverage, loguru, pdoc, pytest, python-dotenv) scraping (Scrapy, Beautiful Soup)
JS: Angular, Vue, htmx, Selenium
various: Java (Spring), PHP

DATA

SQL, Mongo aggregation framework
dbms: Postgres, Mongo, Oracle, MySQL/MariaDB, DuckDB
messaging: RabbitMQ, Qpid, Kafka, ARQ
tooling: visidata, litecli https://github.com/dbcli/litecli/pull/217
dataframes: Polars, Pandas

open source

https://github.com/lusingander/serie/issues/53

technical writer: Stack Overflow (threads vs. processes, debugging, Docker Compose inheritance), technical article covered by Python Bytes podcast
code contributor: The Hitchhiker's Guide to Python (PRs on loguru, Clint), various (Portray, CPython, ptpython, bandit, fff)

EXPERIENCE

🏀 Kero Sports

🗓️ 2025.10-present

pregame resolution

https://bitbucket.org/kerogaming/rush_ml_v2/pull-requests/4844

lookup error happens (per Sentry) on market_info.get("form_actionNumber")
nothing has changed in the code around this LOC for 2+ years
which makes me think this is a data issue, not a code issue
all the issues seem to be with pre-game markets
pregame markets don't have PBP data
pregame markets can (will occasionally? always? never?) have null actionNumber
when the game starts, we attempt to resolve all markets, incl. pregame
if pregame markets lack actionNumber, they can't resolve
so why is this only happening for select pregame markets (e.g. P_4_23_5, P_4_29_4)?
we seemingly handle pregame market state mgmt via get_pregame_publish_market, which should be picking up P_4_23_5, P_4_29_4, and so forth
these markets have ownership of their own unpublish logic (unlike other pregame markets?? "There should be one - and preferably only one - obvious way to do it.")

class P_4_23_5(v.TeamNextOffensivePossessionResult):
    ...
    @classmethod
    def resolve_market(cls, event_info, market, last_row, con: BasketballRmgWorkDB):
        ...
        if last_row.get(
            "@matchstatus"
        ) != BetRadarMatchStatus.NOT_STARTED and not market.get("unpublished"):
            res = ResolveOption.UNPUBLISH

so the issue could be that when we try to resolve them via BasketballRmgResolving.run(), we're trying to lookup form_actionNumber before their own internal unpublish mechanism resolve_market gets called, which means we'll hit this lookup error, which means we'll never resolve them.
Ok, so here's a potential fix for this https://bitbucket.org/kerogaming/rush_ml_v2/branch/fix/basketball/pregame-unpublish-temp-fix?dest=feat_common_service
The fix looks idiotically simple, I know, but think it will handle these deviant markets for now. Mid/long-term we should align them with their fellow pregame markets and move ownership of their resolution outside of themselves.

stack

FastAPI
Sentry
Mongo, Postgres
Kafka, ARQ

🟥 Capp

🗓️ 2024.08-2025.04

project: ERP
role: data eng
contribution: set up EDI connection for suppliers, wrote product category tooling, modernized price updates (Polars, sqlite-utils, visidata)

💿 United Masters

🗓️ 2022.01-2023.12

project: payments
role: backend (Flask, Postgres, Mongo, Kafka)
contribution: projects include hardening data ingregrity for Split Pay system, integrating Chargebee for subscriptions, moving from 3rd party royalty system to in-house replacement

🏦 Eliassen (BNY, Capital One)