Chris Stefan
Senior AI Engineer · Agentic AI in production · EU Citizen · Remote-ready

I build AI in production
for institutional finance.

AI Engineer at PSP Investments (~$300B CAD AUM). I lead development of the Virtual Analyst Platform: agentic AI for 100+ analysts across all asset classes, cutting research cycles ~10× and powering every AI workflow the team ships.

// about
Who I am

I started in Computer Science, moved into Finance & Business Analytics at McGill, and found my edge where the two converge. I joined PSP Investments (first in information security, then digital innovation) where I quickly moved from evaluating alternative data and cyber risk to actually building the AI systems that investment teams depend on. The team evolved from Digital Innovation into AlphaScience, and I converted to full-time AI Engineer in 2023.

Today I'm leading the development of the Virtual Analyst Platform, PSP's agentic AI system in production, used by 100+ analysts across all asset classes. I designed the entire stack from scratch: a custom dual-panel UI with full agentic transparency and source traceability, a workflow launcher for named structured analyses, and the underlying AlphaScience SDK that powers every AI workflow, RAG index, and dashboard the team ships.

Outside of work, I'm a builder. I'm developing digithings.ai, a modular agentic AI toolkit that distils what I've learned at PSP into reusable components anyone can use. Having brought the flagship platform to production and scale, I'm relocating to Europe and seeking new challenges, leveraging my experience of shipping apps in production at scale with other teams across new domains and industries. I want to keep building things that genuinely matter.

// projects
What I've built
Systems in production I designed, built, and own end-to-end.
Virtual Analyst Platform · PSP Investments · All asset classes · Showcased at Databricks Data & AI Summit

PSP's flagship agentic AI platform (300 total users, 100 active & recurring), embedded in analyst–PM daily workflows across equities, fixed income, real assets, and private equity. AI findings flow directly to PMs who audit & red-team theses in the same session.

300
Total users
100
Active & recurring
Lead
Developer
  • Custom dual-panel Gradio UI designed from scratch: chat interface with full agentic step transparency + workflow launcher for named structured analyses
  • Workflow launcher: company overviews, deep research, thesis memos, bull/bear reviews, and buy-side research review
  • Source traceability: every claim links to its exact document, page, and passage, surfaced directly in the UI so analysts never leave the platform to verify
  • Structured data integration: model manipulates datasets, generates charts, and builds fully interactive exportable dashboards on the fly within the chat session
  • Multi-source RAG: broker research, financial filings, earnings call transcripts, and custom financial data indexes
  • Recursive agentic workflow engine with dynamic tool-calling, sandboxed code execution, and automated LLM-driven reporting
  • All powered by the proprietary AlphaScience SDK
Agentic AI SDK RAG OpenAI Azure AI Search Gradio Databricks Python Azure Key Vault
AlphaScience SDK · PSP Investments · Team-wide foundation

A proprietary Python SDK that is the foundation for every AI deliverable the AlphaScience team ships: every LLM workflow, RAG index, and analytical dashboard runs on it.

  • Auth & credentials: unified Azure Key Vault integration and API auth across all services
  • API wrappers: standardized interfaces for OpenAI, Azure AI Search, and Databricks
  • Agent orchestration: reusable abstractions for LLM agents and agentic tool-calling
  • RAG index management: consistent patterns for building, querying, and updating vector indexes
  • Onboarded interns to independently ship features in production using the SDK
Python SDK LLM Agents RAG Indexes Team Standard
Earnings Release Prediction Engine · PSP Investments · Digital Innovation

An ML pipeline predicting the beat/miss likelihood of earnings releases for covered companies, with a full client-facing Power BI dashboard delivering daily-updated predictions.

  • Drove the business case and secured stakeholder buy-in through rigorous POC analysis
  • Point-in-time data pipeline: designed to avoid look-ahead bias using financial domain knowledge
  • Feature engineering grounded in financial fundamentals; full interpretability layer
  • Power BI dashboard with daily beat/miss predictions across all covered companies
Machine Learning Point-in-Time Data Feature Engineering Power BI Python
// experience
Where I've worked
For more details, see the CV
5 years building AI in production at the intersection of institutional finance and agentic systems. EU citizen with right to work in the EU; no sponsorship required. Seeking senior, lead, or principal roles to build high-impact AI products.
Senior AI Engineer · Alpha Science · PSP Investments · Montreal, QC Sep 2022 – Present
  • AlphaScience SDK: full-stack base-class framework; Databricks + OpenAI wrappers; one-liner for agents & vector indexes
  • Virtual Analyst Platform: chatbot UI, step-level tracing, source traceability, workflow launcher
  • Agile roadmap: weekly-evolving priorities; rapid integration of bleeding-edge AI into production
  • Cost savings: in-house platform displaces enterprise tooling at scale
  • Leadership: C-suite collaboration; mentored interns; Databricks Summit demo
Agentic AI RAG Orchestration Azure AI Evals Observability
Analyst · Digital Innovation · PSP Investments · Montreal, QC Sep 2021 – Aug 2022
  • Earnings Prediction Engine: ML beat/miss pipeline, point-in-time data, Power BI for $8B portfolio
  • Alt-data pipeline: feature engineering; stakeholder buy-in across asset classes
Quant Strategy Alt Data ML Pipelines Power BI scikit-learn
Intern · Information Security · PSP Investments · Montreal, QC May 2021 – Aug 2021
  • Enterprise risk framework; automated risk assessment updates
Cybersecurity Risk Analysis
Business Analyst · Lunch à Porter · Montreal, QC Jun 2020 – Jun 2021
  • SEO & A/B testing (2× conversion); inventory analytics for merchandising
SEO A/B Testing Shopify Lightspeed Analytics
// skills
What I work with
Complete skill set across AI in production, data, and finance.
🧠

Gen AI & LLMs

OpenAI API Gemini API Claude API Hugging Face Prompt Engineering Vector DBs Agentic AI RAG Orchestration Evals Observability LangGraph LiteLLM MCP OpenClaw

Data & Cloud

Azure GCP Databricks Spark Delta Lake PostgreSQL SQL Supabase MLOps CI/CD Docker ML Pipelines Power BI
💻

Programming

Python JavaScript HTML CSS FastAPI Pydantic REST APIs SDK Design Git
📊

Visualization

Plotly Dash Matplotlib ECharts Mermaid Tableau
🤖

AI Dev Tools

Cursor Antigravity GitHub Copilot Claude Code Bolt.new Open WebUI Ollama
💰

Quant Trading

QuantConnect IBKR API Nautilus Trader TA-Lib Quant Strategy Alt Data
📄

Data & Research

LSEG Worldscope IBES SEC Filings Broker Research Transcripts scikit-learn Cybersecurity Risk Analysis
🛒

Business Analytics & E-commerce

SEO A/B Testing Shopify Lightspeed Analytics
// interview Q&A
In my own words
Q "Tell me about yourself."

I'm an AI engineer with a background in CS and finance. I've spent the past 5 years at PSP Investments building AI in production for institutional investment teams, leading the agentic AI platform as it grew from POC to production. Full stack ownership: SDK, UI, onboarding, and roadmap. Now I'm looking to expand into new domains and keep building systems that deliver real value.

Q "Why AI? Why finance + AI specifically?"

I came at it from both sides. CS gave me the engineering foundation; finance gave me the domain context to understand what analysts actually need. That mix has been useful for building tools that people trust and use day to day.

  • Finance is demanding for AI: high stakes, messy data, sophisticated users who need to verify everything
  • Working in that environment has pushed me to focus on reliability, traceability, and making outputs verifiable
  • I'm interested in applying those patterns to other industries where the stakes and impact are high
Q "What's the biggest technical challenge you've solved?"

Source traceability. Analysts can't take an AI's word for something. I built a system that links every claim back to its exact document, page, and passage, surfaced in the UI so they can verify without leaving the platform. The engineering was one part; the other was realizing that without that trust layer, the tool wouldn't get used at all.

Structured data integration. Financial datasets are too large for context windows. I built a backend that enables the model to reference tables by path, work with compact summaries, and run ML analysis on top of that layer. Results flow back through the LLM. The same pain points exist in healthcare, legal, and other domains with large structured data.

Q "Where do you see yourself in 5 years?"

I'm passionate about building AI systems in production that matter, and I want to keep learning and growing my expertise in the field. I'm open to any form of remote work; what matters to me is the work itself, the team, the value we ship, and the impact it has on clients.

// side projects & open source
What I build on my own time
Production knowledge from institutional AI, distilled into open-source components anyone can use.
🤖
digithings.ai

Fully modular multi-agent AI toolkit. LangGraph orchestration, LiteLLM universal LLM wrapper, pluggable RAG backends, built-in evals, Docker deployment, MCP & Open WebUI integration, OpenClaw layer for autonomous agent deployment. Direction: “hedge fund in a box”: merging agentic AI with quantitative finance.

  • LangGraph orchestration: reusable agent graph patterns for multi-step agentic reasoning and workflow automation
  • Search engines with index connections: pluggable components supporting Azure AI Search, local search, and OpenAI vector search
  • Universal LLM wrapper via LiteLLM: single interface for any model: OpenAI, Anthropic, Gemini, xAI, Llama, and any local or API-based LLM
  • Config-based Docker deployment: deploy the full stack from a single config file; containers spin up with the correct configuration automatically
  • MCP connection: accessible via Model Context Protocol, connectable to any compatible chatbot or AI client
  • Open WebUI integration: open-source UI for testing, demoing, and interacting with the full toolkit
View on GitHub ↗
// education & certifications
Background
McGill University
B.Com · Finance & Business Analytics
Desautels · 2019–2022 · Montreal, QC
Champlain College
DEC · Computer Science & Mathematics
2017–2019 · Honor Roll · Montreal, QC
🏅
Microsoft Azure AI Fundamentals
Sep 2023 · Credential ID: D9F91E8DF4C7D0CE
English
Mother tongue
French
Native
Romanian
Native
Spanish
Basic
Italian
Basic
// beyond work
What drives me outside the office
The passions, interests, and projects that shape who I am beyond the code.
🛠️
Builder Mentality
I don't just tinker; I ship. From AI toolkits to open-source components to full product builds, every interest becomes a project. I use AI to accelerate everything I make.
📈
Markets & Finance
Actively follow tech, AI, equities, and crypto. My finance background keeps me close to the markets, and my engineering instinct makes me want to automate every thesis and strategy I come across.
🚀
AI Latest Developments
Closely following the frontier: new model releases from OpenAI and Anthropic, the evolution of agentic frameworks, and what's actually shifting in how AI systems are built and deployed at scale.