The next phase of AI won't be defined just by access to AI tools. It will be defined by the talent: Forward-Deployed Engineers, who can sit inside an organization, understand the work, and turn AI capability into meaningful change. Like large corporations, nonprofits, government agencies, small businesses, and community-serving institutions need that talent.
Flood Voice is one such example: a community-powered flood intelligence system designed to help prevent tragedies by reaching the New Yorkers that current emergency systems often miss. Six Pursuit AI-Native Builders – Erica Rowe-Owen, Ethan Davey, Jessenia Cintron, Josue Villalona, Kelvin Saldana, and Shanell Holback – built it over a weekend at Columbia's Narrative 2 Numbers Mixed Methods Hackathon in October 2025, winning Best Use of Open Source Data. In December 2025, they were invited to present at the Columbia Preparedness and Recovery Institute's quarterly meeting. In May 2026, they returned for the PRI Innovation Forum Demo Expo and won Most Community-Minded Innovation. Erica also spoke as part of a Student Innovation panel. The team is now fundraising to launch a six-month pilot across Queens neighborhoods.
Ninety-three percent of the buildings damaged in NYC during Hurricane Ida were outside FEMA's flood zones. (1) Fourteen people died that night. (2) Most were trapped in basement apartments in Queens. Most were Asian and foreign-born. (2) They lived in neighborhoods the federal map classified as minimal hazard.
The storm was forecast. The data existed. NYC's 2018 stormwater drainage study had already mapped the neighborhoods where the drainage system, designed for 1.75 inches of water per hour, would fail. (3) Ida dumped 3.15 inches in a single hour, the wettest on record. (4) FloodNet sensors were already in the ground. Notify NYC was already pushing alerts. Community-based organizations were already on the block, in the languages residents spoke.
The pieces were there. They weren't talking to each other. The deployment layer didn't exist.
That deployment layer is Flood Voice. It runs on two parallel tracks when a flood warning is issued.
The first is community intelligence. The system pulls geolocated flood reports from social media in real time, uses AI to score each one for tone and urgency, and correlates the signal with NYC's FloodNet sensors. The NYC Department of Environmental Protection installed those sensors across the five boroughs to measure water rising on the sidewalk in real time.
The second is direct welfare verification. Automated voice calls and SMS go to vulnerable residents in affected ZIP codes. A single keypress (1 for safe, 2 for need help) escalates to emergency contacts and first responders with the resident's address and medical needs. Target speed: under five minutes.
A Triple Validation Framework cross-checks community signal, FloodNet sensor reading, and government data from FEMA and DOHMH before any alert is acted on. The target accuracy is 85% or higher.
The dashboard is a command center for community-based organizations (CBOs), community health workers, and first responders. Intake forms, currently available in English, Spanish, Bengali, and Mandarin, enable CBOs to register residents into ZIP code pods, with additional languages coming soon. Residents themselves never see the dashboard. They only experience Flood Voice as a single inbound call or text.
In October 2025, six Pursuit Builders signed up for the Narrative 2 Numbers Mixed Methods Hackathon at Columbia University. Dr. Analee Etheredge, Senior Research Scientist at NYC Department of Health, had seen Erica's L2 build, given her feedback, and invited her to the hackathon. Erica showed up with five other Builders.
The hackathon was organized around NYC's 2024 Jurisdictional Risk Assessment, which identified ten public-health hazards New Yorkers face. (5) They were chemical threats, coastal storms, cyberattacks, flooding, hazardous materials releases, infectious disease outbreaks, mass casualty incidents, pandemics, temperature extremes, and utility interruptions. The team picked flooding, which NYC Health had named the most pressing of the ten just months earlier. "Because of the deaths in Hurricane Ida," Erica explains, "we felt that it could have been avoided." A comprehensive emergency alert system for the city's most vulnerable didn't exist.
Over the weekend, they built Flood Voice. They won Best Use of Open Source Data.
The roles weren't preassigned. They emerged from what each Builder brought. Ethan took climate research and infrastructure vulnerability. Kelvin built the data pipelines. Josue owned the design for accessibility. Jessenia led the grant narrative. Shanell shaped the public story. Erica led the public-health framework.
None of them could have built Flood Voice alone. Each Builder brought a different discipline, and each one made the others better in the act of building. This is Many-to-Many at work: nobody was purely a teacher or purely a learner. Knowledge moved in every direction. The team taught each other and became more than what each Builder could have done alone.
Erica has a master's in Population Health Informatics from CUNY School of Public Health. Before she was a Pursuit Builder, she spent over a decade in healthcare access and public health equity.
Most people picture public health as doctors and hospitals. It's actually the work of tracking patterns across thousands of people early enough to stop a crisis. "When COVID hit," Erica explains, "officials weren't waiting for doctors to call in cases. They were watching wastewater, hospital admissions, and geographic clusters, triangulating signals to find where the problem was worst."
Flood Voice does that for flooding. "That's not a tech feature," Erica says. "That's epidemiology."
The reframe from tech build to public health happened organically and intentionally as they presented the product to the public. At the December 2025 PRI quarterly and the May 2026 Demo Expo, "nobody was asking us about the user journey," Erica recalls. "Nobody asked us about the tech stack." Public health asks different questions: what's the evidence, what's the data showing, what's the health impact? Insights like this don't come from a classroom or building in a silo. They come from building, sharing with the public, learning from the feedback, and going back to the drawing board to improve the product.
The city has an estimated 50,000+ illegal basement and cellar apartments housing at least 100,000 people, mostly working-class immigrants and communities of color. (6) Flood Voice is designed for them, and for everyone existing alert systems miss: elderly New Yorkers, people with disabilities, non-English speakers. None of them is going to register directly with a stranger. The trust isn't there.
Community-based organizations already have it. They have the languages and the relationships with the elderly neighbor, the dialysis patient, the family doubled up in the basement. So Flood Voice was built around them.
"We are not coming in as outsiders with a technology solution," the team's partnership invitation reads. "We are coming as neighbors, asking for your help to protect the people you already serve."
No data Flood Voice collects is shared with ICE, NYPD, or any agency outside emergency flood response. That promise is written into every partnership.
"Stop designing for users and start designing for trust," Erica told the PRI Innovation Forum panel. "Scale without trust is just reach. And reach without relationships saves no one."
At the May 2026 Demo Expo, Flood Voice won Most Community-Minded Innovation. Designing for trust before designing for users isn't an instinct from a traditional tech pipeline. It comes from a career that's accumulated other expertise along the way: public health, community work, immigrant advocacy, the lived experience of being the person the system wasn't designed to reach.
The team is currently fundraising for a 6-month pilot to prove its impact across Queens. They're targeting 1,000 registered vulnerable residents, 10 signed partnership agreements with community-based organizations, and a Triple Validation accuracy rate of 85% or higher.
Flood Voice is already proof of something the AI industry is only beginning to name: AI Deployment. The industry calls it Forward-Deployed Engineering, and large enterprises are racing to staff it. Pursuit is training the same workforce for the small businesses, nonprofits, and community-serving institutions enterprise often overlooks.
Our Builders are uniquely equipped for this role. They combine AI fluency with lived experience, domain expertise from non-traditional careers, and a deep understanding of the communities most systems fail to reach. That proximity allows them to build AI solutions grounded in trust, relevance, and real human need.
(1) NYC Office of Management and Budget. NYC CDBG-DR Action Plan: Post-Tropical Cyclone Ida. Technical Amendment 2. October 2024. https://www.nyc.gov/assets/cdbgdr/documents/amendments/Ida_Amendments/NYC_Hurricane_Ida_Action_Plan-TA2.pdf
(2) NYC Department of Health and Mental Hygiene. "Immediate Injury Deaths Related to the Remnants From Hurricane Ida in New York City, September 1–2, 2021." Disaster Medicine and Public Health Preparedness, Cambridge University Press, April 2024. https://pubmed.ncbi.nlm.nih.gov/38577778/
(3) NYC Mayor's Office of Resiliency. NYC Stormwater Resiliency Plan.https://www.nyc.gov/assets/orr/pdf/publications/stormwater-resiliency-plan.pdf
(4) Bowen S, Ortiz LE, et al. "Increasing extreme hourly precipitation risk for New York City after Hurricane Ida." Nature Scientific Reports, 2024. https://www.nature.com/articles/s41598-024-78704-9
(5) NYC Department of Health and Mental Hygiene. "NYC Health Identifies Top Public Health Hazards for New Yorkers." Press release, 2025. https://www.nyc.gov/site/doh/about/press/pr2025/nyc-health-identifies-top-public-health-hazards-new-yorkers.page
(6) Office of the New York City Comptroller. Bringing Basement Apartments Into the Light.https://comptroller.nyc.gov/reports/bringing-basement-apartments-into-the-light/