Start a project
Real Estate Dealroom

Deal flow, underwriting, and investor-ready sharing

Real Estate Dealroom

A portal to manage listings, comps, docs, and pipeline updates with permissioned access.

Timeline

4-8 weeks

Category

Real estate ops

Stack

Next.js, Postgres, maps, file storage, permissions

Delivered by

Lowkey Tech

Type

Sample case study

Overview

A deal platform that centralizes data, documents, and decisions for faster closes.

Findings

  • Local RAG lab indexes dealroom data and uploads for cited retrieval.

  • Search and chat surface grounding metrics: score, source diversity, citation coverage.

  • Answers flag unsupported claims to keep responses audit-ready.

Notes

  • Runs offline with a JSON vector store; OpenAI is optional for richer answers.

  • Uploads support txt, md/markdown, and json files.

  • Re-indexing is safe and replaces stale chunks.

What it does

A standalone Dealroom demo with a local RAG lab that indexes dataset briefs and uploads, then answers with citations and evaluation checks.

Screenshots

RAG lab overview with ingest controls and diagnostics
Search and chat workflow with cited sources
Evaluation mode showing grounding checks

Workflow

End-to-end RAG workflow inside the Dealroom demo.

Step 01

Indexing

Ingest + validate

Seed the dataset pack or upload docs (txt, md, json) with metadata.

Inputs

Seed pack, uploads, metadata

Checks

File type, size, missing fields

Output

Raw text payloads

Step 02

Indexing

Normalize + chunk

Clean text and slice into 350-word segments for retrieval.

Processing

Normalize, dedupe, chunk

Config

350 words, 40 overlap

Output

Searchable text segments

Step 03

Indexing

Embed + index

Generate embeddings and store vectors in a local JSON index.

Model

Local fallback or OpenAI embeddings

Store

JSON vector index

Output

Vector + metadata pairs

Step 04

Query

Retrieve

Top-k semantic search returns the most relevant chunks.

Query

Semantic search + filters

Top-k

5 chunks per request

Output

Ranked evidence set

Step 05

Query

Answer + cite

LLM drafts the response and attaches citations for each claim.

Guardrails

Citation coverage enforcement

Output

Answer + citation map

Fallback

Sources only if unsupported

Step 06

Query

Evaluate

Grounding score, citation coverage, and unsupported claim flags.

Metrics

Grounding, coverage, diversity

Flags

Unsupported claims

Output

Pass/fail signals

What we built

Listings ingestion and normalization with search + filters

Deal pipeline with notes, tasks, and ownership

Underwriting snapshots + comps workflow

Investor/agent portal with permissioned sharing

Document vault with versioning and link-based access

Value delivered

Reduce back-and-forth by keeping everything in one place

Make underwriting repeatable with consistent templates

Share updates safely with investors and partners

Move faster with clearer ownership and next steps

Outcomes

What this delivers in the real world.

Decision speed

Fewer bottlenecks with a shared deal view

Repeatability

Standard underwriting snapshots

Sharing

Permissioned investor portal

Docs

Versioned vault + secure links

Demo

See Real Estate Dealroom in action.

Get a tailored walkthrough for your team.

Request a walkthrough
Real Estate Dealroom | Lowkey Tech