Designing and deploying multi-agent systems for investment research, M&A due diligence screening, and FP&A automation using Python and frameworks like LangChain or CrewAI
Setting up and managing a private cloud environment (AWS or Azure) where all AI workloads run ensuring client data never leaves Tristone's controlled environment
Building and maintaining an internal knowledge base (vector database) that indexes all past project work and makes it searchable by AI agents and analysts
Integrating AI agents with external financial data sources using controlled API connections with rate limiting and audit logging
Owning the prompt library: write, test, version-control, and continuously improve the prompts used by all agents
Data Privacy & Security
Architect a private LLM deployment self-hosted open-source models or private cloud instances so client data never passes through shared public APIs
Implementing data classification: tag incoming data by sensitivity and enforce routing rules so confidential data only reaches approved private models
Maintaining full audit logs of every AI agent action what went in, what model processed it, what came out, who accessed it
Enforcing role-based access controls so analysts use AI tools within defined permissions
Produce monthly privacy compliance reports for senior leadership
Coordinating with External Development Partners
Serving as Tristone's technical point of contact for any outsourced developers or firms building AI modules
Reviewing all code from external partners before deployment checking security, data handling, and quality
Writing technical specifications that translate analyst workflow requirements into developer briefs
Managing all code in a private GitHub repository with documentation Tristone always owns and controls the codebase
Continuous Improvement & Fine-Tuning
Monitoring AI agent outputs daily: track accuracy, hallucination rates, and analyst feedback then improve prompts and configurations
Fine-tuning models on Tristone-specific financial language over time starting with prompt engineering, progressing to RAG, and eventually supervised fine-tuning
Evaluating new AI tools monthly and recommend adoption where there is clear ROI
Training & Supporting the Analyst Team
Running onboarding sessions for analysts on how to use AI tools effectively practical, not theoretical
Building a simple internal dashboard showing team AI usage and time saved per person
Being the go-to person when an agent produces a wrong output diagnose it and fix it quickly
