At Abacum, I own complex financial-modelling and dataset systems end to end, from reconciliation and dataframe execution to public API foundations. I reduced a conflict-ranking workload from roughly two minutes to 20 ms by replacing row-wise processing with columnar execution.
I've extended Abacum's financial-modelling DSL across grammar, validation, autocomplete, dependency graphs, execution graphs, and dataframe operators. I also reduced formula-computation memory pressure by projecting required columns down to S3 reads rather than loading full datasets.
I improve production reliability through Celery and Redis pipeline design, named queues, monitoring, and watchdogs for stuck jobs. I led scalability work for an ML classifier with 14K+ features and moved training and inference to workers so 10GB+ models stayed off API-serving nodes.
Previously, I helped Blacklane's pricing engine support geographic zone-based pricing, roughly doubling revenue in affected markets. I've also built high-throughput Django services, developer tooling, CI/CD pipelines, Kubernetes infrastructure, and ETL systems across product and client work.
