Automated Deal Screening System for Affordable Housing Acquisitions

Summary We're a California-based affordable housing developer (LIHTC acquisition-rehab, preservation, and new construction) looking for a developer or small firm to build a fully automated deal-screening system for our acquisitions team. We screen 2,000+ properties a year against state-specific scoring criteria. Today it is done by hand. An analyst types an address into a form, measures distances to eleven amenity types in Google Maps, checks four other websites for flood zone, rural status, and rent limits, adds the points up manually, picks the total from a dropdown, then re-types the result into our Zoho CRM. It takes about 45 minutes per property. We want that replaced by a system where the input is an address (plus, where available, the seller due diligence file) and the output is a scored, recorded deal in Zoho with a full audit trail. No intake form. No analyst keying in answers. The only human step is reviewing and confirming what the system found, or overriding it with a note. Scope 1. Input. A single property address, or a batch of them. Optionally a folder of seller due diligence documents attached to that property. 2. Automated location scoring. Distance to the eleven amenity categories used in California TCAC site amenity scoring (grocery, transit, public schools, medical, parks, pharmacy, library, and the rest), measured the way the QAP defines them, including the different urban, suburban, and rural distance thresholds. Plus FEMA flood zone, rural set-aside designation, QCT and DDA status, opportunity area map designation, and the applicable rent and income limits. 3. Document parsing. Read seller-provided DD (offering memorandum, rent roll, T-12, HAP contract, existing regulatory agreement, LIHTC documents, PCA) and extract the screening-relevant facts: unit mix, in-place rents, project-based voucher and HAP coverage, existing affordability restrictions and their expiration, year built and last rehab, occupancy. Flag low-confidence extractions for human review rather than guessing. 4. Scoring and outcome. Apply our full criteria set (every question, conditional path, disqualifier, and point value) and produce the score and pass/fail outcome automatically. The manual tally and the second-set-of-eyes re-check both go away. 5. Criteria stored as data, not hardcoded. California scoring rules are revised every year and we operate in several other states. We need to change a threshold, add a question, or load another state’s ruleset ourselves, without commissioning another build. 6. Self-validation. Our current form has rules that were written but quietly stopped firing, and there was no way to see it from inside the tool. The system should check its own ruleset for orphaned or unreachable logic every time criteria change, and flag it. 7. Zoho CRM write-back. Results land in our Zoho Leads module automatically (score, outcome, extracted data, and the record of which lookup returned what). Zoho stays our system of record for the rest of the pipeline. 8. Historical re-scoring. Re-run our roughly 3,000 completed screens through the corrected ruleset and report which deals would have been judged differently. What we provide • Our complete California screening criteria, the live Google Form it currently runs in, and a Canva workflow board mapping the full screen • Roughly 3,000 completed historical screens • Zoho CRM access and the current broker reply templates • The CA QAP and TCAC regulations we score from, and confirmation of which version we work from • About two hours up front with our analyst lead, then roughly an hour a week Tech stack Your call. Tell us what you would build it on and why, where it runs, and what the monthly run cost looks like (hosting, mapping and data APIs, any LLM usage) at our volume of roughly 2,000 screens a year. Who we are looking for • Demonstrated experience automating a rules-based scoring or eligibility process end to end, not workflow glue between SaaS tools • Geospatial and mapping API work (Google Maps or equivalent), plus public data sources such as FEMA, HUD, and Census • Document extraction from messy real-world PDFs and scans, with a real answer for accuracy, confidence thresholds, and what happens when extraction fails • CRM integration experience; Zoho specifically is a plus What to include in your proposal 9. Timeline, and the earliest point our team could start screening real deals against it. 10. Your proposed stack and hosting, with estimated monthly running cost. Ongoing work Additional states after California, plus criteria updates at least annually as new QAPs and scoring rules are published, would likely result in repeat work.