Open to opportunities

Vahan Hovhannisyan

@vahanhovhannisyan

Message

Senior AI/ML engineer building production LLM and agentic AI for financial intelligence.

Zimbabwe

Message

What I'm looking for

I’m looking to build production LLM/agent systems for regulated or high-stakes domains—focusing on reliable extraction, RAG and evaluation, privacy-first architectures, and audit-ready outputs that measurably speed up real analyst workflows.

I’m a Senior AI/ML Engineer with 6+ years building production LLM and agentic AI systems for financial intelligence, delivering enterprise-scale outcomes for clients including Moody’s and Fitch Ratings.

I design multi-model reasoning pipelines, agent orchestration, RAG architectures, and LLM evaluation frameworks—at scale (100K+ documents/month). My work spans credit, securities, regulatory filings, and structured/unstructured financial data, with an emphasis on measurable quality, auditability, and lineage.

I’ve founded two live fintech document AI products. As AuditFlow’s sole engineer, I built structured extraction with 99% field-level accuracy and end-to-end lineage reconstruction; as part of the Optimus Securities Platform, I’m developing production-ready securities operations workflows for banks and investment firms.

Experience

Work history, roles, and key accomplishments

Current

Co-Founder, Optimus

Current

Optimus Securities Platform

Jan 2025 - Present (1 year 5 months)

Designed and built a production-ready securities operations platform covering the full trade lifecycle from order management through FIX execution, SWIFT settlement, custody, corporate actions, and client reporting.

Securities Operations Order Management Fixed Swift Corporate Actions Data Modeling Production Systems

Current

Founder, AuditFlow

Current

AuditFlow

Jan 2024 - Present (2 years 5 months)

Designed and built a live AI platform for structured extraction from bank statements and IRS Form 1120s, achieving 99% field-level accuracy with full data lineage reconstruction. Architected end-to-end document processing pipelines using LLMs, OCR, and VLMs to handle varied multi-page layouts and edge cases.

LLM Structured Extraction Data Lineage Document Processing Pipelines Python AWS Document Automation

Data Science Lead

Cognaize

Aug 2020 - Apr 2025 (4 years 8 months)

Architected multi-model LLM and agentic AI pipelines with model routing, tool-calling, and structured output schemas, achieving 88% field-level accuracy across complex financial document workflows. Increased analyst throughput 3x via a human-in-the-loop branching system and delivered audit-ready, lineage-preserving outputs for Moody’s and privacy-first, jurisdiction-local processing for Fitch Rati

Agentic AI Multi Model Routing Structured Outputs Human In The Loop

Machine Learning Engineer

SoftConstruct

Jun 2019 - Aug 2020 (1 year 2 months)

Developed a question-answering NLP system for a production chatbot and a supervised ML-based bot detection system. Researched reinforcement learning and game theory approaches for strategic AI agents.

NLP Chatbot Bot Detection Reinforcement Learning Game Theory Python Model Training