Senior AI/ML Engineer

I lead a team of four AI engineers, define model architecture, conduct code and prompt reviews, and own delivery across concurrent financial document intelligence client pipelines. I also drive technical pre-sales through enterprise banking demonstrations and proof-of-concept presentations.

I own a production document automation pipeline processing more than 100,000 financial documents per month, reducing manual review time by 70%. I architected self-correcting agentic LLM workflows, financial statement hierarchy inference, and multi-model extraction pipelines achieving more than 90% field-level accuracy.

I built end-to-end machine learning pipelines for multilingual financial document parsing using commercial and open-source LLMs, supporting more than 20 languages and over 50 financial institutions. I designed RAG systems using LangChain and FAISS, along with agent-based postprocessing for statement balancing and error detection.

I re-architected PDF extraction from a single large model into task-specific small language models, reducing GPU compute time by 47%, inference time by 64%, and GPU memory requirements from 80 GB to 12 GB. I also developed human-in-the-loop routing, privacy-first localized processing, a Python SDK for enterprise clients, and improved OCR and table detection for low-quality scanned documents.

I analyzed client financial statements, including balance sheets, income statements, and cash flow statements, across IFRS, US GAAP, and local GAAP frameworks.

I identified reporting discrepancies and collaborated with the data science team to define financial data schemas and labeling strategies for early automated financial extraction models.

I monitored real-time transaction data for fraud patterns across payment cards, ATMs, and POS terminals.

I produced analytical fraud reports to support risk assessment activities and regulatory requirements.

Bachelor's degree in Economics and Trade.