Machine Learning Engineer

About RealQuantRealQuant is building thefirst vertical AI platform for commercial real estate— turning OMs, rent rolls, and financials into structured insights that power underwriting, reporting, and portfolio intelligence.We're redefining how deals move fromOM ? LOI in under 30 minutesby combining institutional real estate expertise with modern AI automation.Role OverviewYou'll own theAI enginebehind our document parsers, data feeds, and underwriting agent —training and fine-tuning multimodal modelsin Azure AI Foundry to deliversub-10s parsingand automated deal outputs (LOIs, pitch decks, summaries, and recommended assumptions).Tech StackCore AI Platform:Azure AI Foundry (multimodal) · Azure ML (training + fine-tune) · Azure OpenAIRetrieval & Data:Azure AI Search (vector + hybrid) · pgvector · Databricks (labeling + ETL) · Azure Postgres Flexible · Blob StorageAPIs & Workflows:FastAPI inference services · Azure Durable Functions · Event Grid (ingest triggers) · Service Bus (optional) · API Management (APIM)Ops & Monitoring:App Insights / Monitor · Key Vault (Managed Identity) · GitHub Actions CI/CDWhat You'll BuildTrain and fine-tune LLMs + embeddings for OM, Rent Roll, and P&L parsing.Detect and learn document layout patterns to reduce human-in-loop (HITL) review.Build labeled datasets and evaluation benchmarks in Databricks + Azure ML.Integrate 3rd-party market data and create retrieval pipelines via Azure AI Search / pgvector.Develop theAI Underwriting Agentthat generates deal summaries, LOIs, pitch decks, and assumptions.Deploy models via Azure ML / Durable Functions with scalable APIs.Optimize inference pipelines forsub-10s end-to-end latency, including preprocessing and write-back to Postgres.Track accuracy, recall, latency & cost with App Insights dashboards; iterate rapidly using Cursor / Claude Code + HITL QA.Expected Outcomes3+ production-ready parsers (OM, Rent Roll, or P&L) deployed in Azure100+ labeled examples + versioned evaluation datasets=95% extraction accuracy vs manual baselineAI-generated deal summary & LOI draft live in Excel / webConfidence-scored mappings for Chart of Accounts & unit-type labels with HITL reviewParsing pipeline optimized forsub-10s total latencyExperience5–7 yrs ML engineering (modeling, evaluation, MLOps)3+ yrs LLM / Document AI (fine-tuning, embeddings, RAG)Hands-on with Azure AI Foundry / Azure ML / OpenAI APIPython + PyTorch / Transformers expertiseExperience with pgvector or AI Search for retrievalFinTech, PropTech, or Data SaaS background a plusFluent English · ownership mindset · outcome-drivenSelection Process90-minute async coding / model taskTechnical interviewPaid trial sprint (build a real parser module)Compensation & SetupIndependent Contractor |Hourly DOEFull-time remote · EST overlap · Immediate startLong-term contract with growth potentialHow to ApplySend yourGitHub, demo, or portfolioof embedding / LLM work, plus a short note on a model you've taken from prototype ? production.