← dealpath / Senior Product Manager, AI
brief / art_f8SrD2499uM
Company snapshot
Dealpath is a purpose-built real estate investment management SaaS platform that has powered over $10 trillion in transactions for hundreds of firms, including Blackstone, Nuveen, LaSalle, CBRE IM, and MetLife. The company is backed by a strong strategic investor syndicate including Blackstone, Nasdaq, 8VC, JLL Spark, and Morgan Stanley Expansion Capital, giving it both institutional credibility and deep distribution into the real estate investment community. Dealpath is currently in an active AI acceleration phase, seeking to embed AI across the full deal lifecycle — from sourcing and screening through underwriting and asset management. The engineering and product culture appears to value hands-on technical PMs who can prototype, build evals, and engage directly with ML engineers. Specific recent product launches or internal engineering blog details are not publicly available; inferences above are based on the JD and public company positioning.
Team stack
Based on the JD and public signals: core platform is likely enterprise SaaS (web-based, likely React or similar front-end); AI/ML layer is actively being built out with LLMs, RAG pipelines, and agentic systems (likely OpenAI/Anthropic APIs plus internal fine-tuning, based on JD language); MCP is explicitly called out as a known technology; vector store and retrieval infrastructure likely in place or being built (specific vendor unknown — Pinecone, Weaviate, or pgvector are common in this space); back-end likely Python-heavy given ML focus; data layer likely includes proprietary deal and asset data used for model training and RAG grounding; CI/CD and eval pipelines are explicitly desired but likely early-stage. All stack inferences beyond MCP and LLM/RAG are uncertain and based on JD language.
Likely questions (10)
| area | question | why |
|---|---|---|
| system_design | Walk us through how you would design a RAG pipeline to help an underwriter at Blackstone extract key deal metrics from a 200-page offering memorandum. What are the failure modes and how do you build evals around them? | JD explicitly requires working knowledge of RAG pipelines, AI evals, and diagnosing model failures; underwriting is called out as a core deal lifecycle stage. |
| domain | What are the highest-leverage AI opportunities across the real estate deal lifecycle — sourcing, screening, underwriting, and asset management — and how would you prioritize which to build first at Dealpath? | JD asks the PM to 'identify the highest-leverage opportunities to apply AI across the deal lifecycle'; this tests domain judgment and strategic prioritization. |
| coding | Describe a time you vibe-coded or rapidly prototyped an AI proof-of-concept to test with users before committing to a full build. What did you build, how long did it take, and what did you learn? | JD explicitly calls out 'comfortable spinning up rough AI prototypes and vibe coding proof-of-concepts' as a required skill. |
| behavioral | Tell me about a time you had to manage enterprise client expectations during an AI calibration phase when the model wasn't yet performing reliably. How did you maintain trust? | JD calls out 'engage directly with sophisticated enterprise clients, set realistic expectations during AI calibration phases, and maintain trust even when the product is still evolving.' |
| system_design | How would you design a human-in-the-loop feedback system for an AI feature that extracts cap rates and NOI from deal documents? When do you automate, and when do you require human review? | JD explicitly requires 'experience building AI products that require human calibration, oversight, and feedback loops — and the product intuition to know when to automate and when to keep humans in the loop.' |
| domain | How would you build and maintain an AI eval suite for a document extraction feature used by investment professionals? What metrics matter, and how do you detect regression? | JD requires 'creating and iterating on AI evals, diagnosing model failures, and driving continuous improvement' as a core PM skill. |
| behavioral | Describe a product you took from 0 to 1 — what was your process for defining the roadmap, aligning stakeholders, and deciding what to cut? | JD asks for 'lead the AI product roadmap — balancing near-term wins with long-term platform thinking'; 0-to-1 experience is implicitly required. |
| culture | Dealpath's clients include some of the most sophisticated real estate investors in the world. How do you approach thought leadership and external representation of a product that is still evolving? | JD explicitly asks the PM to 'host webinars, contribute to thought leadership content, or speak at industry events' and be the external voice of AI strategy. |
| domain | What is MCP (Model Context Protocol) and how would you use it to extend Dealpath's AI capabilities? Can you give a concrete example of an MCP server you would build for a real estate investment workflow? | JD explicitly lists MCP as a required area of working knowledge; this is a differentiating technical signal for the role. |
| behavioral | Tell me about a time you had to make a build-vs-buy decision for an AI capability. What framework did you use and what was the outcome? | JD requires 'making informed build-vs-buy decisions' as a core competency for the AI PM role. |
Talking points
- Built a production RAG retrieval pipeline with ChromaDB, multi-provider LLM orchestration (Claude, GPT-4, Gemini) with fallback routing, and structured output validation at Fintellect AI — directly analogous to the document intelligence and deal data extraction workflows Dealpath needs. Also built Redfin/Zillow-integrated real estate CMA agents in StreamIO, demonstrating real estate domain fluency with live data APIs.
- Shipped aeval, a local-first AI model evaluation platform with 5 eval types (factuality, reasoning, instruction-following, safety, code generation), bootstrap confidence intervals, Welch's t-test, Cohen's d effect size, saturation detection, and CI/CD regression gates — directly matching Dealpath's requirement for a PM who can build and maintain AI evals and diagnose model failures.
- Implemented OpenClaw multi-agent orchestration framework with gateway protocol, subagent delegation, and session management in StreamIO — and benchmarked 12 RL algorithms across TRL, VeRL, OpenRLHF, and NeMo RL in the RL Workbench — demonstrating the agentic systems and MCP working knowledge the JD explicitly requires (StreamIO also ships an MCP server exposing screen capture tools).
- At Intuit, delivered the ICE Self-Service platform that reduced developer onboarding from 2–3 weeks to minutes, scaled throughput from 6K to 50K TPS, and drove 675M+ engagements — proving the ability to own a technical platform roadmap, align cross-functional stakeholders, and deliver measurable business outcomes at enterprise scale, which maps directly to Dealpath's need for a PM who can execute end-to-end.
- Holds a California Real Estate Broker/Sales Agent license and has built real estate-specific AI agents (CMA reports via Redfin/Zillow APIs) in production — providing genuine domain credibility when engaging with Dealpath's Blackstone, Nuveen, and CBRE clients, and a concrete foundation for identifying high-leverage AI opportunities across the deal lifecycle.