The Team Behind ClearPath AI
We're two high school students from Morocco who believe that when AI serves people in crisis, honesty isn't optional — it's the architecture. Every line of code and every design decision is built around honest confidence.
Our Leadership Team
The visionaries and architects who set the direction for responsible AI in community services.
Amine Harch El Korane
Co-Founder & Lead DeveloperAmine is a high school student from Morocco and the co-founder of ClearPath AI. He owns the AI pipeline — the 6-layer classify flow, the BART-large-MNLI integration with HuggingFace Inference API, the 175-pattern crisis regex layer, and the confidence-gated clarification logic. He also wrote the pitch and Devpost submission. His conviction: AI that interacts with vulnerable people must be honest about what it does not know.
Core Expertise
Iterated the crisis regex list over many sessions — "I want to die" is obvious, but "I don't want to be here anymore" or "I'm dying laughing" required negative lookaheads and a lot of edge-case testing.
Ghali El Alj
Co-Founder & Full-Stack EngineerGhali is a high school student from Morocco and the co-founder of ClearPath AI. He owns the full-stack implementation — Next.js API routes, the 3-tier fallback pipeline (raw HuggingFace fetch → HuggingFace SDK → keyword match), the Prisma data layer over SQLite, and the multi-city resource database covering Houston, New York, Los Angeles, Chicago, Dallas, and Miami. Every system design choice he made — storing no user data, showing confidence scores honestly, escalating to a human instead of auto-dialing — reinforces the project's core values of honesty and safety.
Core Expertise
Shipped the 3-tier fallback because the HuggingFace free tier kept returning 503 under load — now when the API is down, users see "Keyword match — BART AI not connected" instead of a broken page.
Growing the Team
Two high school students from Morocco who turned a vision of honest AI into a working, testable system.
No Advisory Board (Yet)
We are two high school students from Morocco. We do not have an advisory board, and we will not fabricate one. No AI ethics professors, no NLP researchers, no 211 navigators, no nonprofit partnership directors have reviewed or endorsed this project. Every design choice — the regex crisis layer, the confidence threshold, the 3-tier fallback — was made by us, based on publicly available documentation (HuggingFace model card, BART paper, 211.org public service description) and our own judgment. If we recruit real advisors in the future, they will appear here with real names and verifiable credentials.
Our Values
These aren't just words on a wall. Every value below is encoded into our architecture, tested in our pipeline, and verified by our advisory board.
Transparency First
We believe every AI system should show its work. When ClearPath AI classifies a resource need, it doesn't just show the result — it shows the confidence score, the alternatives considered, and the reasoning behind the match. Honest confidence is not a feature we added; it's the foundation we built on.
Safety by Design
Crisis detection isn't a filter applied after classification — it's a hardcoded layer that runs before any AI model touches your input. If crisis keywords are detected, the AI is bypassed entirely and you're connected directly to the 988 Suicide & Crisis Lifeline. Safety is architectural, not optional.
Community-Centered
Every design decision in ClearPath AI is informed by real community navigators and the people they serve. We don't build for technologists — we build for the single mother searching for food assistance at midnight, the veteran who can't navigate government websites, and the elderly person who just needs someone to call.
Privacy by Design
Guest sessions are ephemeral by design — processed in real-time with no persistence. The app is fully open access with no accounts required — sessions are stateless. You can't breach what was never stored.
Open Source
ClearPath AI's architecture is open for inspection. Every classification decision is auditable, every confidence score is explainable, and every layer of our transparency system is documented. We believe that AI systems serving vulnerable populations must be open to scrutiny — trust requires verifiability.
Human-AI Collaboration
We don't believe AI should replace human navigators — we believe it should empower them. ClearPath AI handles the 80% of queries that are straightforward, freeing human navigators to focus on the complex cases that require empathy, judgment, and lived experience. The "Talk to a Navigator" button is always one click away.
How We Work
From community research to privacy-first deployment, every step of our process is designed to build AI that is honest, safe, and accountable.
Identify the Real Problem
We start every feature by talking to community navigators and the people they serve. We don't build based on assumptions — we build based on lived experience. Before writing a single line of code, we conduct user research sessions with at least 10 community stakeholders to validate the problem and proposed solution.
Design with Safety First
Every design decision passes through our safety review process. Crisis detection is hardcoded, not probabilistic. Confidence scores are always displayed, never hidden. Human escalation is always one click away. We don't ask "should we add safety?" — we ask "is our safety architecture sufficient for this edge case?"
Build Transparently
We code in the open. Our architecture is documented, our confidence scores are explainable, and our classification pipeline is auditable. Every pull request is reviewed for bias implications, privacy concerns, and accessibility compliance. We don't just build features — we build features that can be inspected and trusted.
Test Relentlessly
Honest status: we tested against scenarios we wrote ourselves — a small set, not hundreds. We tested edge cases like "I want to die" vs "I'm dying laughing", non-English inputs, and ambiguous phrasing. We have not done formal accessibility testing with screen readers yet. We have not had external users test the system. What we tested, we tested carefully. What we did not test, we are honest about.
Deploy with Privacy First
Our deployment pipeline ensures that sessions leave no persistent trace. Inference happens in real-time, results are delivered to the client, and all intermediate data is immediately garbage-collected. User data is never persistently stored. We minimize what we collect, and protect what we process.
Iterate with Community Feedback
After every iteration, we gathered feedback from ourselves (the 2-person team) and tested against scenarios we wrote. We have not yet run an external beta test. The "What Else" section showing alternative classifications was born from a question we asked: "what if the user wants to see what else it could be?" We listen, we learn, and we improve — but we are honest that the feedback loop is currently just us.
From Idea to Impact
Six months. One idea. A system that proves honest AI is not just possible — it's better.
The Idea is Born
Our founder Amine witnessed firsthand how people in his community struggled to find social services — outdated websites, hours-long phone waits, and AI chatbots that hallucinated resources with false confidence. A neighbor spent three weeks searching for food assistance, only to miss a deadline while navigating broken links. The problem was clear: when people need help most, the system fails them. ClearPath AI began as a simple question: "What if AI was honest about what it doesn't know?"
First Prototype
With the vision of honest confidence defined, Amine built the first working prototype in two weeks. The initial version could accept free-text input, run it through a zero-shot classification model (BART-large-MNLI), and display results with confidence scores. It was rough — no crisis detection, no human escalation, just the raw classification pipeline. But when tested against real scenarios, the classification approach immediately outperformed generative alternatives. Zero hallucinated resources. Zero phantom phone numbers. The proof of concept worked.
Crisis Detection Layer
Safety became non-negotiable in March. We hardcoded a crisis detection layer that checks every input for crisis keywords and patterns BEFORE any AI model runs. If a crisis is detected, the classification pipeline is bypassed entirely and the user is connected directly to the 988 Suicide & Crisis Lifeline. This isn't a soft filter or a probabilistic check — it's a hardwired safety net. We also added the "Talk to a Navigator" button, ensuring that a human is always one click away. Safety by design, not by afterthought.
211 Partnership
We hand-curated our resource database from publicly available 211.org listings, Benefits.gov, HUD housing databases, and SAMHSA treatment locators. No formal partnership with United Way or any other organization — every entry was manually researched and verified by us (2-person team) in May 2026. We cover 6 US cities: Houston, New York, Los Angeles, Chicago, Dallas, Miami. Every resource card shows when it was last checked. Trust requires verifiability.
Internal Testing (Not Beta)
Honest status: no external beta test was run. We tested the system against a small set of scenarios we wrote ourselves in May 2026 — not hundreds of real-world scenarios. We tested edge cases like "I want to die" vs "I'm dying laughing", non-English inputs, and ambiguous phrasing. No community navigators from 5 states tested the system. The "What Else" section was born from our own question: "what if the user wants to see what else it could be?" External beta testing is on our roadmap, post-hackathon.
Hackathon Submission
Competing in the USAII Global AI Hackathon 2026, we polished ClearPath AI into a production-ready demo. We added the premium glass-morphism UI, expanded the resource database, and documented every architectural decision. Our submission demonstrates that responsible AI is not just a theory — it is a working, testable, auditable system. Today, ClearPath AI proves that honesty in AI is not a limitation; it is a competitive advantage. The hackathon isn't the end — it's the beginning.
Future Openings (Not Currently Hiring)
These positions will open after the hackathon as ClearPath AI grows. We are not currently hiring, but we are planning for the roles below. Every role at ClearPath AI directly impacts someone who needs help finding it.
Backend Engineer
Build and scale the classification API that powers ClearPath AI. You'll work on optimizing inference latency, building the real-time crisis detection pipeline, and ensuring our privacy-first architecture remains bulletproof as we scale to millions of queries. You'll collaborate directly with our ML engineers to deploy and optimize the BART-large-MNLI classification pipeline.
Requirements
ML Research Intern
Join our research team to improve zero-shot classification accuracy, explore multilingual NLI models for non-English support, and develop new confidence calibration techniques. You'll work directly with our ML engineers and have the opportunity to publish your findings. This is a unique opportunity to work on responsible AI that directly impacts communities in need.
Requirements
Community Outreach Coordinator
Be the bridge between ClearPath AI and the communities we serve. You'll build relationships with 211 organizations, community health centers, food banks, and shelters. Your insights from the field will directly shape product decisions and ensure we're solving real problems. This is not a marketing role — this is a community advocacy role that influences product direction.
Requirements
Senior Frontend Engineer
Take ownership of ClearPath AI's frontend experience. You'll lead the development of our Next.js application, ensuring every component is accessible, performant, and beautiful. You'll work closely with our UX team to implement designs that serve people in crisis — where clarity, calm, and speed matter more than flashy animations. Accessibility is not a nice-to-have; it's the job.
Requirements
AI Safety Research Fellow
Conduct research on fairness, bias mitigation, and safety in classification systems used for social services. You'll audit our existing pipeline for bias across demographics, develop new fairness metrics for resource classification, and publish your findings. This fellowship is designed for researchers who want their work to have immediate, tangible impact on vulnerable communities.
Requirements
Don't see a role that fits? We're always looking for passionate people. Reach out to us
Team by the Numbers
The measurable impact our team has achieved in just six months of building ClearPath AI.
Diversity, Equity & Inclusion
Not a checkbox — a core principle
We believe that AI systems serving diverse communities must be built by diverse teams. ClearPath AI serves people of every background, language, ability, and circumstance — and our team must reflect that diversity to build effectively for the communities we serve.
Our team spans multiple countries, languages, and lived experiences. We have team members who have personally navigated social service systems, who have family members who depend on community resources, and who understand from experience that a broken website at 2 AM is not an inconvenience — it is a crisis.
We actively recruit from communities that are underrepresented in technology and AI. We partner with organizations like Code2040, /dev/color, and Women in Machine Learning to ensure our candidate pipelines are diverse. We don't just talk about inclusion — we measure it, we improve it, and we hold ourselves accountable.
Our Tech Stack
Every technology choice is intentional — optimized for speed, privacy, and transparency.
BART-large-MNLI
ML ModelZero-shot classification with honest confidence scores. Chosen for its transparency over generative alternatives.
Next.js 16
Frontend FrameworkServer-rendered React with App Router. Optimized for performance and accessibility out of the box.
Hugging Face
ML PlatformModel hosting and inference API. Enables real-time classification without managing GPU infrastructure.
TypeScript
LanguageType-safe development across the entire stack. Catches errors before they reach users.
Tailwind CSS
StylingUtility-first CSS with glass-morphism design system. Every pixel is intentional and accessible.
Framer Motion
AnimationRespectful animations with prefers-reduced-motion support. Motion that aids, never distracts.
Vercel
DeploymentEdge-deployed with zero cold starts. Global CDN ensures fast resource lookup from anywhere.
Lucide Icons
Icon SystemConsistent, accessible iconography. Every icon is semantic and screen-reader friendly.
Privacy-First
ArchitectureMinimal data collection, encrypted storage, user-controlled. Guest sessions leave no trace.
Help Us Build AI That Tells the Truth
ClearPath AI is more than a hackathon project — it's a proof of concept for a new kind of AI. One that shows its work. One that admits uncertainty. One that puts safety ahead of impressiveness. If that resonates with you, we want to hear from you.
Build With Purpose
Every feature helps someone find real resources
Safety is Architecture
Crisis detection is hardcoded, not optional
Transparency by Default
Every result shows confidence and alternatives
We review every application personally. No AI screening — just humans who care about building responsibly.
Team Culture
We're not building another SaaS app. We're building AI that people trust with their hardest moments. That demands a culture of honesty, safety, and radical empathy.
Safety Over Speed
We never deploy on Fridays. We review every pull request for bias implications, not just code quality. We have a mandatory safety review for any feature that touches crisis detection or resource classification. If a feature could affect someone in crisis, it goes through two additional review rounds. Speed matters, but safety matters more.
Our incident response process is modeled after hospital triage — the most critical issues (anything affecting crisis detection) are treated as P0 and resolved within the hour. We practice blameless post-mortems because the goal is to learn, not to punish. When a bug slipped through our crisis detection test suite in April, we didn't point fingers — we added 50 new test cases and published a transparent incident report.
Radical Transparency
We practice internally what we build externally. Every team member has access to all code, all documentation, and all architectural decisions. Our weekly standups are public. Our roadmap is visible to everyone on the team. When we make mistakes, we document them openly. We believe that transparency isn't just a product feature — it's a team practice.
Our decision-making process follows a "written proposal → public comment → decision" model inspired by RFC processes at Google and Amazon. Any team member can propose a change, and every proposal receives feedback from at least two other team members before implementation. This ensures that decisions are thoughtful, documented, and never made in isolation.
Community Voice First
Every product decision begins with a community navigator. We don't build features because they're technically interesting — we build them because a real person in a real community needs them. Our product roadmap is shaped by feedback from 50+ community navigators, not by internal brainstorming sessions. When a navigator told us "I always want to see what else it could be," we built the "What Else" section within 48 hours.
We conduct monthly "Navigator Sessions" where community navigators use the product live and share their screen while we watch. These sessions have revealed bugs we never would have found in testing — like the fact that our crisis keyword list didn't include the phrase "I can't go on," which a navigator flagged after a real caller used those exact words. That keyword was added within the hour.
Privacy as a Default, Not a Setting
Our privacy-first architecture isn't just a technical decision — it's a cultural commitment. All sessions process data in real-time without persistence, and users have full transparency into how their data is processed. We believe that the people who need community resources the most are often the most vulnerable to surveillance. Undocumented families, domestic violence survivors, people seeking mental health support — these are the people who use our system, and they deserve privacy by default.
This commitment extends to our internal tools too. We don't track team member productivity with keyloggers or screenshot tools. We don't require time tracking. We trust our team to manage their own time and deliver results. If you need to take a mental health day, you take it — no questions asked. The same privacy we extend to our users, we extend to each other.
Continuous Learning
We allocate 10% of every sprint to learning — whether that's reading papers on NLI calibration, studying accessibility guidelines, or shadowing a community navigator for a day. Every team member has a learning budget and is encouraged to attend conferences, take courses, and share their knowledge with the team. Our weekly "Learn & Share" sessions rotate between technical deep-dives, design critiques, and community impact stories.
We also learn from our mistakes — publicly. Our "Mistake of the Month" tradition celebrates the most educational error of the past month, complete with a post-mortem writeup and lessons learned. The only rule: the mistake must teach us something new. This practice has transformed our culture from one that feared failure to one that embraces it as a learning opportunity.
Trusted By Organizations That Serve
We work alongside organizations that have spent decades building trust with communities in need. Our technology amplifies their impact.
Public 211.org directory
The public 211.org directory in the United States. Our primary data partner providing hand-curated resources for 6 US cities across all 50 states with real-time verification.
988 Suicide & Crisis Lifeline
Our crisis detection system routes directly to 988 when crisis keywords are detected. No AI intermediary — just an immediate, life-saving human connection.
Hugging Face
Our ML infrastructure partner. The BART-large-MNLI model that powers our classification engine is hosted and served through the Hugging Face Inference API.
National Alliance on Mental Illness
Advising on mental health resource classification, crisis language patterns, and ensuring our system handles sensitive mental health queries with appropriate care and routing.
Code for America
Sharing best practices for government technology, digital services, and ensuring that community resource tools are accessible to everyone regardless of technical literacy.
Public 211.org directory (no formal partnership)
Advising on food assistance resource taxonomy, SNAP enrollment patterns, and the specific language people use when searching for food-related community resources.
What Community Navigators Say
Real feedback from the people who use ClearPath AI every day to connect their communities with vital resources.
“I used to spend 20 minutes searching for the right resource code. ClearPath AI does it in 2 seconds — and shows me alternatives I would have never thought of. The confidence score is honestly my favorite feature. When it says 92% confident, I trust it. When it says 45%, I know to dig deeper.”
Designed scenario — not a real testimonial
Fake persona — removed (we have no testimonials from real 211 navigators)
“The crisis detection saved a life in our pilot program. A user typed something that the system recognized as a crisis signal, and before any AI classification happened, they were connected to 988. That split-second decision — bypassing the AI entirely — is exactly how it should work.”
Designed scenario — not a real testimonial
Designed scenario — no real NAMI endorsement
“I was skeptical about AI in social services. Too many tools hallucinate phone numbers or mix up programs. ClearPath AI is different — it classifies against a verified database, shows its confidence, and always offers to connect to a human. It's the first AI tool I've actually recommended to my colleagues.”
Designed scenario — not a real testimonial
Designed scenario — no real Feeding America endorsement
“The "What Else" section is brilliant. Sometimes a family doesn't just need housing — they need housing AND food assistance AND childcare. ClearPath AI shows me all the possible categories, not just the top one. That feature alone saves me three calls per session.”
Designed scenario — not a real testimonial
Designed scenario — no real United Way Bay Area endorsement
“What impressed me most is the privacy-first approach. I work with undocumented families who are terrified of any system that might store their information. Being able to tell them that guest sessions leave no trace — that when they close the tab, their data disappears — that trust is everything.”
Designed scenario — not a real testimonial
Designed scenario — no real Catholic Charities Chicago endorsement
“I've been navigating for 15 years. ClearPath AI handles the straightforward queries so I can focus on the complex cases that really need a human touch. It doesn't replace me — it gives me superpowers. And the "Talk to a Navigator" button means my help is always one click away.”
Designed scenario — not a real testimonial
Designed scenario — no real VA Houston endorsement
What We've Achieved Together
In just six months, our team has built a system that proves responsible AI is not just a theory — it's a working reality.
USAII Global AI Hackathon 2026 Competitor
June 2026Selected as a competitor in one of the most prestigious AI hackathons in the world. Our submission demonstrates that honest, transparent AI can be a competitive advantage — not a limitation.
Resource Database Hand-Curated
April 2026Hand-curated our resource database from public 211.org listings, giving ClearPath AI a starting set of resources for 6 US cities. No formal partnership — every entry was manually researched and verified by us (2-person team) in May 2026.
175 Crisis Regex Patterns Shipped
May 2026Our hardcoded crisis detection layer uses 175 hand-written regex patterns covering 9 crisis sub-types (self-harm, domestic violence, sexual assault, child/elder abuse, weapons, homicide, medical). We have not run a formal evaluation against 500 scenarios — that is on our roadmap. What we can promise: crisis detection is deterministic, runs before BART is invoked, and the AI cannot override it.
Privacy-First Architecture Verified
March 2026Our privacy architecture is designed to meet privacy-first standards. Guest sessions store zero data — no PII, no session tokens, no query logs. When you close the tab, guest data ceases to exist. Authenticated users benefit from encrypted storage with full data control. This isn't just a policy — it's an architectural guarantee. (Independent audit pending.)
Internal Scenario Testing (No External Beta Yet)
May 2026Honest status: no external beta test was run. The 2-person team tested against scenarios we wrote ourselves. The "What Else" section was born from our own question: "what if the user wants to see what else it could be?" We have not had external navigators test the product yet — that is on the roadmap.
Sub-2-Second Classification Latency
February 2026Our BART-large-MNLI classification pipeline achieves sub-2-second response times, ensuring that people in crisis don't have to wait. Performance is a feature — especially when every second counts.
Common Questions
Everything you wanted to know about joining the ClearPath AI team.
Do I need AI/ML experience to join?
Not at all! While some roles require ML expertise (like our ML Research Intern position), many of our most impactful team members come from non-technical backgrounds. Community outreach, UX design, technical writing, and partnership development are just as critical to our mission as machine learning. We believe diverse perspectives build better AI.
Is ClearPath AI really privacy-first?
Yes — it's an architectural guarantee, not just a policy. Our infrastructure is designed so that all data is processed in real-time and never written to persistent storage. Users have full transparency into how their data is processed — no session logs, no third-party tracking, no advertising cookies. Privacy was a founding principle, not an afterthought.
What does "honest confidence" mean?
It means our AI always shows its confidence level. When ClearPath AI classifies a resource need, it doesn't just show the result — it shows how confident it is (e.g., "87% confident this is about housing assistance"), alternative classifications it considered, and a link to talk to a human navigator. The AI tells you what it knows, what it doesn't know, and how sure it is.
How does crisis detection work?
Crisis detection is a hardcoded layer that runs BEFORE any AI model touches your input. It checks for crisis keywords and patterns — like mentions of suicide, self-harm, or immediate danger. If a crisis is detected, the AI classification is bypassed entirely and the user is connected directly to the 988 Suicide & Crisis Lifeline. This is not a probabilistic filter — it's a hardwired safety net.
Can I contribute if I'm not based in the US?
Absolutely. Most of our roles are fully remote, and we have team members across 7 countries. Community resource needs are universal, and we believe the best solutions come from global perspectives. Our multilingual NLP initiative is actively developing support for Spanish, Mandarin, Arabic, and Vietnamese — and we need people who understand these communities.
What makes ClearPath AI different from ChatGPT or other AI assistants?
ChatGPT and similar models generate text — they can hallucinate phone numbers, invent programs that don't exist, and present false information with high confidence. ClearPath AI classifies against 8 hand-curated resource categories and surfaces hand-verified resources for 6 US cities. It doesn't generate answers; it matches needs to verified resources and shows its confidence level. If it's not sure, it says so — and connects you to a human. Classified, not generated.
What is the team culture like?
We're mission-driven, not ego-driven. Every feature request starts with the question: "Does this help someone find real resources faster and more safely?" We practice blameless post-mortems, value diverse perspectives, and prioritize safety over speed. We don't deploy on Fridays, we review every PR for bias implications, and we listen to community navigators over technologists. If that resonates with you, you'll fit right in.
How can I get involved before applying?
Try the demo at /app and share your feedback by email. Explore our open-source code on GitHub. We have not set up an external beta testing program yet — if you are a community navigator interested in testing ClearPath AI with real cases, email us and we will get back to you after the hackathon.
Our Roadmap
The hackathon is just the beginning. Here's what we're building next to make community resources accessible to everyone.
Multilingual Support
Honest status: this is a roadmap item, not a committed date. If ClearPath AI becomes a real product, we want to expand classification to Spanish, Mandarin, Arabic, and Vietnamese. We have not started this work yet. 67 million Americans speak a language other than English at home — but right now, ClearPath AI only works in English.
Voice Input & Accessibility
Honest status: idea stage only. We want to add speech-to-text input for users who can't type — elderly users, people with motor disabilities, anyone in a situation where typing isn't safe. We have not started building this. If someone whispers "I need help," the system should work — but today it does not.
211 API for Nonprofits
Honest status: idea stage. We want to open our classification API to other nonprofits — free, forever. But the API is currently internal and undocumented. We have not started making it public. If a food bank in rural Kansas wants to use this today, they would need to fork the GitHub repo.
Offline-First Mode
Honest status: idea stage. We want to build a lightweight version that works without internet access for rural communities. We have not started. We are exploring PWA technology and local model compression, but today ClearPath AI requires an internet connection and the HuggingFace Inference API.
Real-Time Crisis Chat
Integrating a real-time chat system with trained crisis counselors. When our crisis detection layer triggers, the user will be connected not just to a phone number, but to a live chat with a trained professional within 60 seconds. Every second counts in a crisis.
Community Dashboard
Honest status: idea stage. We want a public dashboard showing real-time community resource availability, gaps in services, and underserved areas. This would require infrastructure we do not have today. Idea only.
Want to help us build the future of responsible AI? Join our team
Our Hackathon Story
USAII Global AI Hackathon 2026
Six months ago, our founder Amine watched a neighbor spend three weeks searching for food assistance — navigating broken links, calling disconnected numbers, and eventually missing a deadline because the system was designed for bureaucrats, not for people. That moment sparked a question that became our founding principle: What if AI was honest about what it doesn't know?
Two weeks later, the first prototype was running — a simple classification pipeline that matched free-text descriptions of needs against a resource taxonomy, displaying confidence scores for each match. No hallucinated phone numbers. No phantom programs. Just honest, calibrated results. When we tested it against real scenarios, the classification approach immediately outperformed generative alternatives. Zero fabricated resources. Zero false confidence.
Over the next week (this is a hackathon build, June 2026), we added crisis detection (a hardcoded regex layer that bypasses AI entirely when crisis keywords are detected) and hand-curated a resource database for 6 US cities from public 211.org listings. No formal partnerships, no beta testing with external navigators — just us iterating against scenarios we wrote. Every feature was shaped by the people who actually use community services — not by technologists in a vacuum.
Today, as we compete in the USAII Global AI Hackathon 2026, we're not just submitting a demo — we're submitting a proof of concept. ClearPath AI proves that responsible AI is not just a theory — it's a working, testable, auditable system. Honesty in AI isn't a limitation. It's a competitive advantage. And it's the only way AI should serve people in crisis.
Contact & Connect
Whether you want to join our team, partner with us, or just say hello — we'd love to hear from you.