slides/asce/deck.md

metadata

marp: true
theme: riprap
paginate: true
size: '16:9'
title: >-
  Riprap. Citation-grounded flood-exposure briefings for any place in New York
  City.
description: ASCE NY State Convention, Albany, May 13, 2026

ASCE NY State Convention  ·  Albany, NY  ·  May 13, 2026

Riprap

Citation-grounded flood-exposure briefings for any place in New York City.

Speaker

Adam Munawar Rahman · IBM · MS CE, NYU

Invited by

Andrew Hicks

---

00 · Learning objectives

What you will take away.

After this session, you will be able to:

1. Describe a citation-grounded architecture for synthesizing multi-source flood evidence into auditable, site-specific narratives.
2. Identify where this approach is appropriate (screening, grant evidence, capital planning) and where it is not (hydraulic modeling, stamped deliverables).
3. Evaluate the guarantees and limitations of LLM-based evidence synthesis in civil engineering practice.
4. Apply the Five-Stone architecture to riverine, ice-jam, and dam-failure flooding.

---

01 · The problem

When you assess flood exposure, the evidence sits in eight or more places.

For a capital project, a grant application, a vulnerability assessment, or a property disclosure — the relevant evidence sits across eight or more disconnected primary sources. Synthesizing them into a citable narrative takes hours of GIS work per site.

Federal

FEMA NFHL
USGS 3DEP LiDAR
USGS HWMs (Ida, Sandy)
NOAA CO-OPS tide

State

NPCC4 SLR projections
NYS Mesonet
NWS METAR / watches
NY EJNYC FVI

City

NYC DEP stormwater scenarios
NYC 311 flood complaints
FloodNet sensor network
NYC DOB filings

The gap

No common schema. Different vintages. Different spatial resolutions. Different epistemic tiers.

Each site synthesized by hand.

When a number meets resistance, the only defense is the audit trail.

---

02 · Solution

A flood-exposure briefing for any place in New York City.

---

03 · Architecture

Five Stones. Each with one job.

query → Planner (Granite 4.1 3B, intent classification) → Stone roster → Capstone (Granite 4.1 8B + Mellea) → briefing

Cornerstone · USGS 3DEP 2020

Microtopography (HAND / TWI)

HAND0.82 m TWI14.3 Elev.2.1 m MSL Pct. lower78%

[topo]

Keystone · TerraMind-NYC 2024

Building footprint coverage

48.41%

250 m radius · Buildings LoRA adapter

[keystone_bldg]

Touchstone · NYC 311 live

Flood complaints · 200 m buffer

19 requests · 5-yr lookback

[nyc311]

Lodestone · Granite TTM r2 live

Surge residual nowcast

peak surge residual · 9.6 h horizon

[ttm_surge]

Real evidence cards rendered by the live system  ·  442 East Houston Street, Manhattan.

Capstone Granite 4.1 8B + Mellea rejection sampling  ·  numerics_grounded · no_placeholder_tokens · citations_dense · citations_resolve  ·  reroll until every claim cites its source  →  cited 4-section briefing

---

04 · Demo

Live demo.

“Hollis, Queens”

A neighborhood-scale briefing. NYC DEP and OEM planners use this shape of query when scoping where the next $30B stormwater priority site should land.

5.8 s

end-to-end

4 / 4

grounding checks every run

8+

primary public-record sources

---

05 · Civic applications

The civic case for civil engineers.

Grant evidence

HUD CDBG-DR and FEMA BRIC vulnerability assessments. Riprap auto-generates the per-NTA evidence section for each site in a program area. Citable, reproducible, open-source.

Capital project screening

NYC DEP Bluebelt expansion, NYCHA resilience hardening, MTA station prioritization, DOE school siting. Site-by-site evidence packages at the screening tier, before the hydraulic modeling budget is spent.

NY State Infrastructure Report Card

The 2026 report is in preparation. Riprap is the per-place evidence layer for the flood-exposure chapter of any future NY State infrastructure report — reproducible at every address.

Property disclosure compliance

NY’s March 2024 Property Condition Disclosure flood-risk amendment requires sellers to disclose flood history. Riprap is the citable narrative behind the disclosure — every claim sourced.

---

06 · What Riprap is not.

What Riprap is not.

The civil engineer carries the stamp. Riprap surfaces the evidence the engineer judges.

Not a hydraulic model

Riprap does not replace HEC-RAS, SWMM, or ICM. It synthesizes evidence from completed modeling work; it does not produce new flow or stage estimates. No substitute for a calibrated hydraulic model.

Not a stamped deliverable

The briefing is a starting point for a memo, not the memo itself. Professional judgment, field reconnaissance, and the engineer’s stamp are required for any actionable deliverable.

Not a substitute for site investigation

Microtopography is from 1 m USGS 3DEP LiDAR, appropriate for screening, not for design. Field reconnaissance, soil borings, and survey are not replaced.

Not a risk score

Riprap does not output a 1–10 or 1–100 number. Score-based tools (First Street, ClimateCheck, Jupiter) are different products for different audiences. Riprap is the evidence audit trail behind any such judgment.

---

07 · Directions

Where this goes from here.

The architecture is data-choice-specific, not code-specific.

Upstate NY flooding

The same five-Stone pattern for riverine, ice-jam, and dam-failure flooding. Different primary sources, same architecture.

Historical-event mode

Re-run the system against snapshot data from any past date. Calibration as a core feature.

Stones as standalone packages

Each Stone runs alone. Pull one without the full Riprap stack.

Cross-domain

The same pattern for transit, water, energy, and structural-condition reporting. Flood is the first domain.

---

08 · How it was built

The art of the possible.

Three days of AI-assisted development, on top of months of design thinking. Four foundation models. Three Apache-2.0 NYC fine-tunes trained on AMD MI300X for the AMD × lablab.ai Developer Hackathon (May 4–10, 2026).

Apache-2.0 end-to-end on public-record federal, state, and city data. No commercial APIs contacted at runtime.

Built in three days. Designed over months. The tools have shifted what one engineer can ship.

Foundation models

IBM Granite 4.1 8B (synthesizer) · IBM Granite Embedding 278M (RAG) · GLiNER (typed extraction) · vLLM on AMD MI300X

NYC fine-tunes (Apache-2.0, HF Hub)

Prithvi-EO-2.0-NYC-Pluvial (flood detection, IoU 0.598) · TerraMind-NYC-Adapters (LULC + Buildings) · Granite-TTM-r2-Battery-Surge (surge nowcast, RMSE 0.157 m)

Agentic framework

Burr FSM · Mellea rejection sampling · LiteLLM Router (vLLM / Ollama failover) · FastAPI SSE stream

---

09 · Discussion

What I want from this room.

I am a software engineer, not a civil engineer. The system I just showed you is opinionated about what counts as evidence: citation-grounded, silent when uncertain, public-record only. But I am less sure about where it falls short of how a stamped engineering deliverable would need to behave.

Three questions for the room:

1. Where in your practice would a tool like this be useful, and where would it be a liability?
2. What evidence sources are you using that Riprap does not yet know about?
3. What would have to be true for a citation-grounded narrative tool to be trusted as a screening-tier deliverable?

---

Open-source · Apache-2.0 · github.com/msradam/riprap-nyc

---

Riprap · citation-grounded flood briefings

github.com/msradam/riprap-nyc

---

Apache-2.0 · public data only · IBM Granite 4.1 · AMD MI300X · Mellea grounding

ASCE NY State Convention · Albany, NY · May 13, 2026

Dam mark: “Dam” by Chintuza via the Noun Project, CC-BY 3.0.

---

Appendix A · The receipts

5 of 5 NYC addresses. Every claim verified, every run.

address	intent	wall	steps	verified
442 E Houston St · LES	address	7.6 s	19	4/4
80 Pioneer St · Red Hook	address	13.1 s	19	4/4
100 Gold St · Manhattan	address	11.2 s	19	4/4
Hollis · Queens	neighborhood	5.8 s	9	4/4
Coney Island · Brooklyn	neighborhood	9.9 s	9	4/4

Wall-clock

5.8–13.1 s

vLLM on AMD MI300X

Evidence layers

5

Stones per briefing

Grounding

4 / 4

source checks every run

---

Appendix B · Primary sources

Sources. Every claim traces to one of these.

Federal

FEMA NFHL (current)
USGS 3DEP 1 m LiDAR (2020)
USGS HWMs — Sandy 2012, Ida 2021
NOAA CO-OPS tide gauge, Battery (live)
NWS METAR / flood watches (live)

State / regional

NPCC4 SLR projections (2023)
NY EJNYC Flood Vulnerability Index (2024)
NYS Mesonet (live)
NY Property Condition Disclosure (Mar 2024)

City

NYC DEP stormwater scenarios (2024)
NYC 311 flood complaints (live, 5-yr)
FloodNet sensor network (live)
NYC DOB filings (live)
NYC Open Data — NYCHA, DOE, MTA, hospitals

All datasets are public-record. No commercial data APIs. No proprietary hazard scores.