New York Hackathon
Musa Capital New York Hackathon
Submit your innovative multi-agent AI solutions.
Event Details
Submission Guidelines
Ensure your project aligns with the hackathon themes.
Provide clear and concise descriptions.
Working demos are highly encouraged.
Originality and innovation will be key judging criteria.
Cast Your Vote!
Support your favorite projects by voting.
Vote Now
Voting closes on June 28, 2025.
Project Submission
Fill out the form below to submit your project for the New York Hackathon.
Submissions
All Submitted Projects
Browse the projects teams have entered so far.
ill-think-of-the-name-later
Team: Just Me
Members: Just me.
Problem: Finding parking in the city is a pain in the 🍑
Solution: My solution for the problem statement is to help people find parking in almost all of NYC with using the DOT cameras that run a computer vision model to identify feasible parking options based on the current environment seen by it. Using open data provided by the City to fill in deadzones and show where parking is likely rather than almost guaranteed with the vision system.
We guide you to a parking SPOT without ANY friction.
Cyligo
Team: AJ Adejare
Members: AJ Adejare
Problem: Medications don't necessarily work well under certain heat and health conditions may be worst under worst heat. Given rising temperatures that cause more erratic conditions, we built a prototype for healt.
Solution: Leverage agents to grab informations from APIs and local text while prompting users in a chat to bring better context. Vibe code agent Bolt was used to develop the platform. We call it Cyligo a combination of Cyanea Caligo: "Cyanea" (Latin for a deep blue/cerulean) and "Caligo" (Latin for mist/darkness, evoking a hazard).
MetroVision
Team: Agness Lungu
Members: Esparance Tuyisime, Satyam Chatrola, Parinita Sedai, Satya Kamala Immidisetty
Problem: Navigating New York City's subway system can be confusing and anxiety-inducing, especially for newcomers, tourists, people with cognitive or visual impairments, and even long-time residents during service changes. Traditional map-based tools like Google Maps often fail in underground environments, lack real-time accessibility insights, and don’t capture the spatial complexity of certain subway entrances, exits, or transfers. Many commuters still get lost or take longer routes, costing time and creating daily stress. Additionally, unexpected train delays further disrupt commutes without adaptive rerouting that reflects current conditions.
Solution: Our solution is a mobile app that uses AI to provide video-guided, hyper-personalized subway navigation. Users can receive step-by-step visual directions—like “turn left at the blue sign” or “go down the stairs near the mural”—powered by Llama's vision capabilities to interpret and describe the user’s surroundings in real-time. This makes the experience more intuitive, especially for those with navigation challenges. On the backend, the app predicts train delays using historical MTA data and real-time feeds, dynamically suggesting alternate routes when necessary. Together, these features turn a fragmented and often frustrating journey into a smooth, accessible, and intelligent transit experience for NYC's 4 million daily subway riders
CityZen AI
Team: RSGateways
Members: Roshan Sharma
Problem: New York City drivers waste over 100 hours a year searching for parking, costing them an average of \$2,400 annually in time, fuel, and fines. The lack of real-time, hyperlocal parking data and fragmented information leads to inefficiency, frustration, and aggressive ticketing.
Solution: Your AI agent solution is a real-time, intelligent parking assistant that leverages hyperlocal data, predictive analytics, and user behavior to guide drivers to available legal parking spots, avoid fines, and minimize search time. It integrates city regulations, traffic patterns, and curbside data to deliver personalized, enforcement-aware parking recommendations—saving time, money, and stress.
LeaseLink
Team: LeaseLink Team
Members: Purabh Singh, Ayush Chadha
Problem: As of April 2025, over 350,000 New Yorkers were without homes, highlighting a severe housing accessibility crisis in one of the most densely populated cities in the world. Our solution to this problem, LeaseLink: Tinder for Housing. Despite abundant rental listings online, finding suitable housing remains inefficient, overwhelming and depressing. Conventional websites like Craigslist and Zillow are overflowing with useless or out-of-date listings, lack personalization, and frequently do not effectively match landlords and tenants. Both parties lose out on chances as a result of this disconnect, which also prolongs the housing search. By rethinking the rental experience through a swipe-based, match-driven interface, SwipeHousing solves this issue and makes the process quicker, easier, and more entertaining for both landlords and renters. Now, with LeaseLink, you can find your dream home while doomscrolling on a Thursday afternoon. No more endless scrolling, awkward emails, or ghost listings. With LeaseLink, renting feels less like a chore and more like a game.
Solution: The AI agent in LeaseLink works as an intelligent property recommender that personalizes the browsing experience for each buyer. When a user creates an account, the AI analyzes their profile and preferences like budget, locations, bedroom count, and income/credit score. It then combines this with their past swiping behavior to understand what exactly the user is looking for. The application sends all this data to OpenAI's GPT-4o model, which scores each available property from 1-100 based on how well it matches the user's needs and past choices. Properties are then sorted by these AI-generated scores, so users see the most relevant listings first rather than random properties. This means someone looking for luxury apartments will see high-end properties, while budget-conscious users get affordable options prioritized. Using this AI agent, we seek to make finding housing as personalized and efficient as possible.
CarePet NYC
Team: Heal n' Howl
Members: Akaash Mohan Saxena, Aditi Hebbar
Problem: In urban environments like New York City, many individuals struggle to manage their physical and emotional health due to high stress levels, poor lifestyle choices, and the overwhelming healthcare system. Over 30% of adults in NYC report feeling stressed regularly, and 60% of individuals have trouble navigating healthcare services, leading to deteriorating well-being and untreated chronic conditions.
While wearables capture physiological data, they lack personalized, real-time insights, and do not address mental health concerns. Emotional well-being is often overlooked, and behavioral nudges for better habits are scarce. Additionally, 40% of individuals report losing or misplacing critical health documents like prescriptions or Medicaid letters, making it harder to stay on top of health management.
Solution: CarePet is an AI-powered behavioral health animated companion designed to tackle these challenges by combining multimodal data (e.g., facial expressions, mood journaling, wearables) to offer real-time, personalized insights. Using remote photoplethysmography (rPPG) and emotion classification, CarePet estimates heart rate, breathing rate, and emotional states. The AI-driven agent provides proactive nudges, such as reminders to hydrate or exercise, helping users improve their habits and reduce stress. By integrating local health service recommendations and document scanning, CarePet ensures that users can access essential health resources when needed. This solution targets a 30% increase in user health engagement and a 20% improvement in emotional well-being within 3 months of use.
CivicMind
Team: Dawn
Members: Brian Lovejoy, Dawn Sullivan
Problem:
Urban systems struggle to understand the emotional and cognitive states of their residents.
In high-stress environments like New York City, decisions made by commuters, city workers, and vulnerable populations are often shaped by invisible stressors—ranging from overburdened transit routes and noise pollution to procedural injustice and social isolation.
Yet current civic technologies fail to sense, interpret, or respond to these affective patterns.
Without real-time cognitive feedback, city services remain reactive, fragmented, and emotionally blind—limiting their ability to improve resilience, equity, and quality of life for 8.5 million people.
Solution: CivicMind is a multi-agent AI system that senses, interprets, and responds to emotional stress in urban environments. It integrates data from quantum-inspired wearable sensors and NYC Open Data to map the cognitive-emotional landscape of the city in real time.
The Sensing Agent captures biometric stress indicators using photonic quantum sensors embedded in wearables.
The Insight Agent analyzes public datasets such as 311 calls, noise complaints, and transit delays to identify environmental stressors.
The Feedback Orchestrator cross-references physiological and civic data to deliver actionable insights to city agencies, enabling adaptive, emotionally responsive services.
Finally, the Visualization Agent translates affective signals into intuitive dashboards for policy decisions. By revealing how residents feel—not just what they report
—CivicMind empowers NYC to make smarter, faster, and more equitable decisions.
RenterOS
Team: Renters
Members: Levern Currie
Problem: The project aims to solve the fragmentation and inefficiency of the urban rental market for tenants. Renters face a disconnected and often overwhelming ecosystem for managing their housing journey. Key problems include:
Difficulty in Accessing Affordable Housing: Navigating complex and time-sensitive housing lotteries and voucher programs is a significant barrier for many.
Disjointed Rental Management: Tenants must use separate platforms for making payments, requesting repairs, storing documents, and communicating with landlords, leading to inefficiency and poor record-keeping.
Lack of Transparency and Fairness: The rental process can be opaque, with tenants facing potential bias, unresolved housing violations, and difficulty holding landlords accountable.
Challenges in Co-Living: Finding compatible and reliable roommates is a stressful and often unsuccessful process.
Inability to Build Credit: On-time rent payments, often a household's largest expense, typically do not contribute to building a tenant's credit history.
RentOS addresses this by consolidating these disparate functions into a single, tenant-centric platform.
Solution: Our solution is a multi-agent AI system where specialized agents work in concert to create a seamless and intelligent rental experience, effectively acting as a personal assistant for each tenant.
Lottery & Voucher Agents: These agents proactively monitor housing databases, identify lotteries and vouchers for which the user is eligible based on their profile and income data, and automate the application process through the "Lottery Autopilot" and "VoucherLink" features.
Co-living Agent: This matchmaking agent analyzes user profiles, lifestyle preferences, and budget constraints to suggest highly compatible roommates, facilitating the creation of trusted co-applicant groups.
Compliance & Repairs Agents: The Compliance Agent helps users document and report housing violations and instances of bias, automatically generating formatted complaints for bodies like the HPD. The Repairs Agent streamlines the maintenance process by triaging requests, coordinating with property management, and tracking resolutions.
Wallet & Credit Agent: This agent manages all rental-related finances, automates payments through the "Rent Wallet," and, most importantly, reports on-time payments to credit bureaus, allowing tenants to build their credit score through regular rent payments.
By integrating these specialized agents, the platform transforms the rental lifecycle from a series of stressful, disconnected tasks into a unified, automated, and empowering experience.
What the Rent
Team: What the Rent
Members: Krishiv Seth, Snow Kang, Sameer Jadhav, Sanchit Bansal
Problem: The sticker price of rent on websites like StreetEasy and Zillow only reveal a part of the cost. What about the hidden costs? High utility bills, unsafe commutes, and noisy nights all add up; these hidden costs should be easily accessible to New York residents when choosing a place to live. Our solution is What the Rent, a Chrome extension on top of StreetEasy that reveals these details for renters/buyers to make a truly informed decision.
Solution: We use an orchestrated multi-agent system that, given a residential address, activates specialized agents to uncover hidden living costs:
1) The Energy Agent leverages New York’s publicly available Energy Consumption by Building dataset to estimate average monthly utility costs for that specific address.
2) The Safe Routes Agent analyzes common routes to frequented destinations like work or school and recommends the safest paths based on local crime and transit data.
Together, these agents provide renters and buyers with a richer, data-driven understanding of a home’s true cost and livability, enabling smarter, safer housing decisions in New York City.
Gradma Ui
Team: Gradma Ui
Members: Shoha, Hubert, Aleks, Sam, Geordie
Problem: The project, "311 Portal Navigator for Seniors," addresses a critical accessibility and usability problem for a specific demographic: tech-non-literate elderly users who need to interact with online government portals like NYC's 311 service to submit complaints.
Solution: The project proposes an intelligent GUI overlay application that acts as a personalized, AI-powered digital assistant, guiding the user through the 311 portal step-by-step.
CommuteSense
Team: CityGPT
Members: Philip
Problem: Project Name: “CityGPT”
Problem Statement: “Navigating NYC’s 40+ agencies and complex bureaucracy wastes hours for residents seeking basic services. CityGPT is an AI assistant that instantly connects New Yorkers to the right department, forms, and processes, streamlining civic engagement for 8.3 million residents.”
Solution: CityGPT is an AI chatbot that serves as a single point of entry for all NYC government services. Using natural language processing, it understands resident requests in plain English and provides step-by-step guidance, required documents, and direct links to relevant forms. The AI learns from 311 data and city websites to provide accurate, up-to-date information and can even pre-fill forms with user data.
Interview Ready
Team: Rockers
Members: Thet Mon Aye, Shaishav Pidadi
Problem: Interview processes are often inefficient, inconsistent, and lack actionable feedback for both candidates and interviewers.
Most tools either don’t provide real-time guidance during interviews, fail to adapt questions based on previous answers, or leave candidates without useful post-interview insights. Traditional solutions are time-consuming and don’t scale for modern, remote, or high-volume hiring needs.
Solution: Our Interview App uses an AI-powered agent to simulate an interactive interviewer—adapting questions in real-time based on candidate responses (contextual memory), handling both chat and voice modes, and guiding the flow naturally.
After the interview, the AI agent generates an in-depth analysis: highlighting strengths, improvement areas, and fit, and makes this report instantly downloadable as a PDF.
This automates, personalizes, and streamlines the interview process—delivering consistent, actionable feedback and reducing interviewer workload.
TextMyCommute NYC
Team: CommuteNYC
Members: Sebastian Manalo
Problem: Commuting in a major metropolitan area like NYC is fraught with unpredictable delays, service changes, and disruptions across multiple transit systems (MTA, NJ Transit, etc.). Existing apps and alert systems are often generic, overwhelming, or require constant checking. Commuters need a simple, accessible, and intelligent tool that tells them about issues that specifically impact their journey, and provides grounded alternatives when problems arise.
Solution: To create a personalized, proactive, and intelligent transit information service for New York City and metro area commuters. By leveraging a Retrieval-Augmented Generation (RAG) model and real-time transit data, the product will reduce commute-related stress and save time by providing highly relevant and actionable disruption alerts and on-demand query responses via SMS.
Breaking Barriers: Democratizing Education
Team: Degrowth
Members: Alessandra Serrano, Dean Hu
Problem: Educational Equity Crisis: NYC’s Department of Education serves 1.1 million students across 1,700+ schools, but navigating this vast system is a privilege reserved for those with time, English fluency, and insider knowledge.
Real Impact Stories: - A Spanish-speaking mother in the Bronx keeps her daughter in an underperforming zoned school because she doesn’t know about the arts magnet program 15 minutes away - An 8th grader in Queens with coding talent never learns about the specialized STEM high schools because guidance counselors are overwhelmed - A family new to the US settles for the closest school, unaware their child qualifies for ESL support programs in better-rated schools nearby
The Core Inequity: College students can customize their entire educational experience, but K-12 families—especially immigrant and non-English speaking families—are trapped by zip codes and information barriers.
Solution: Multilingual educational equity platform that uses voice recognition and Google Translate to help immigrant families navigate NYC's complex school system in their native languages (Spanish, Arabic, Chinese, etc.), with AI agents processing natural language queries to match families with appropriate schools based on their specific needs. The platform was deployed using Manus and demonstrates how multi-agent AI systems can democratize access to educational opportunities by breaking down language barriers and simplifying the school discovery process for non-English speaking families.
Cheap Eats NYC
Team: Cheap EATZZZ
Members: Connor Fata
Problem: New York City's food scene is a paradox of abundance and inaccessibility. While the city boasts over 200,000 restaurants, finding genuinely affordable meals under $12 has become a desperate daily struggle for millions of residents and visitors. Students survive on ramen, families skip meals to save money, and tourists get trapped in overpriced tourist traps while authentic $5 pizza slices and $8 halal carts remain hidden in plain sight. Traditional food apps prioritize expensive restaurants and charge crushing delivery fees, creating a digital divide that locks out the very people who need affordable food most. In a city where rent consumes 50%+ of income, the inability to efficiently discover budget-friendly eats isn't just inconvenient—it's a crisis affecting public health, local business survival, and economic equity across all five boroughs.
Solution: Cheap Eats NYC deploys a sophisticated multi-agent AI system powered by Perplexity API and built with Cursor AI to democratize food discovery in America's most expensive city. The solution orchestrates multiple intelligent agents:
1. **Perplexity Search Agent**: Leverages real-time web search capabilities to discover hidden gems, food truck locations, and daily specials that traditional databases miss
2. **Data Fusion Agent**: Combines NYC Open Data (200K+ restaurants), Perplexity insights, and social signals to create comprehensive restaurant profiles
3. **Geospatial Intelligence Agent**: Uses advanced geocoding and distance algorithms to map food deserts and optimize search results by proximity
4. **Price Intelligence Agent**: Analyzes menu data, reviews, and social media to accurately predict meal costs and identify true budget options
5. **Recommendation Engine**: Learns from search patterns to surface personalized cheap eats based on user behavior and preferences
Built entirely with Cursor AI's code generation capabilities, the React application features real-time search, interactive mapping, and mobile-first design. The AI agents work collaboratively to process complex queries like "vegetarian food under $10 near Columbia University" and return actionable results in seconds, transforming how 8.3 million New Yorkers discover their next affordable meal.
"Taste Not Waste"
Team: Jocelyn and Pranav
Members: Jocelyn Marie Goode, Pranav Kowadkar
Problem: In 2017, NYC did a great thing! It passed a law promising to provide FREE school lunch to all of its 1.1 million students enrolled in K-12 public schools. However, this promise produced a great problem: 80 million pounds of food waste and food-soiled paper waste.
How might we reduce the amount of waste and save NYC money, while promoting a culture of respect for food?
Our solution will enable food waste prevention and food waste reduction by a moderate estimate of 15% in NYC schools, which currently amounts to 80 million pounds of waste annually, costing the City of New York $1.7 billion dollars annually.
It will also raise awareness and educate students about the value of food and encourage a culture of food respect.
Through participation from students and school administrators, the platform we have developed will reduce food waste and food-soiled paper waste by 12 million pounds, annually saving NYC $255 million dollars a year
Solution: We used Gemini to help trouble shoot the concept and reveal ethical and logistical problems. Next we had Gemini create a prompt for Bolt to create code and built our platform solution. We also create a variation of the solution using Manus. Our subsequent steps included refining the UI on each page until it satisfactorily represented most of the main components.
Civvy
Team: Civvy
Members: uinytpon
Nana YaaSerwaah, Elena, Ashanti, Q[,h qo
Problem: Many New Yorkers struggle to cut through partisan chatter and patch together reliable facts about local candidates, and there’s little reward for the effort. Civvy lightens that load with a four-question, issue-first onboarding that serves bias-safe content, an AI “Swipe-Flag-Fact” feed that reveals contradictions in real time, and MetroPoints that turn quizzes or vote pledges into OMNY rides and neighborhood perks. Together, these features make staying informed quick, trustworthy, and genuinely rewarding.
Solution: Civvy runs on three core AI agents that map directly to our USPs.
* **Issue-First, Bias-Safe Personalization** – Onboarding answers feed into the *IssueMatcher* model → topic-diversity algorithm injects surprise issues → *FeedCurator* assembles a balanced Story/Reel queue tailored to each user’s ZIP code.
* **Swipe-Flag-Fact Accountability** – Candidate clips flow to the *FactHunter* NLP engine → contradiction search across voting records & news → human-verified flags overlay sources on the post in real time.
* **Play-to-Participate Rewards** – User events (quizzes, verified flags, vote pledges) stream into the *RewardMaster* orchestrator → XP converts to MetroPoints → wallet syncs with OMNY and local-merchant APIs for instant perks.
SafteyGPT
Team: Saftey
Members: Sam Ruan
Problem: Standard navigation apps find the fastest route but ignore real-time, street-level hazards like construction, sanitation issues, or public disturbances.
People walking in cities like NYC have no easy way to know if their path is safe and clear. They can't personalize their route to avoid things that make them feel unsafe or uncomfortable.
SafetyGPT solves this. It combines Google Maps routes with live 311 incident data. Using AI, it scores the safety of a walking path based on user-selected hazards and helps you find a genuinely safer way to your destination.
Solution: The "Map Guy" (get_route.py): Its only job is to get the route map from Google.
The "Problem Finder" (get_hazards.py): It just looks up the list of all recent problems from the city's 311 data.
The "Expert" (our AI in ai_analyzer_local.py): This is the smart one. It looks at each problem and decides how serious it is, giving it a score from 1 to 10.
The "Location Checker" (is_hazard_near_route): It takes the map and the list of problems and checks, "Is this problem actually on our walking path?"
The "Manager" (the main app.py file): It's the boss. It tells everyone what to do in the right order and adds up the final safety score.
Varosync
Team: Varosync, Inc.
Members: Harry Kabodha, Ayman Khaleq
Problem: Unexpected toxicity in clinical trials attributed to poor pharmacology.
Solution: Circadian aware research engine for clinical trials
AI Recruitment Assistant: Multi-Agent Resume Matching System
Team: Rude_monkey
Members: Dhruv chaubey
Problem: The current job matching process is inefficient and often fails to accurately match job seekers with suitable job openings. Traditional job boards and matching services rely heavily on keyword searches and basic filtering, which can lead to irrelevant results and a poor user experience for both job seekers and employers. This inefficiency results in longer job search times for candidates and higher recruitment costs for employers.
Solution: The Multi-Agent AI Resume Matching System solves traditional job matching inefficiencies by employing four specialized AI agents that work together to provide comprehensive analysis. The Resume Parser Agent extracts structured information from resumes, the Job Description Parser Agent identifies requirements and qualifications, the Matching Agent performs intelligent multi-dimensional scoring using semantic similarity algorithms, and the Report Generator Agent creates actionable insights with personalized recommendations. This system delivers precise percentage matches across skills, experience, education, and soft skills, significantly reducing recruitment time and improving candidate-job alignment accuracy compared to basic keyword-based matching systems.
AI Recruitment Assistant: Multi-Agent Resume Matching System
Team: Rude_monkey
Members: Dhruv Chaubey
Problem: The current job matching process is inefficient and often fails to accurately match job seekers with suitable job openings. Traditional job boards and matching services rely heavily on keyword searches and basic filtering, which can lead to irrelevant results and a poor user experience for both job seekers and employers. This inefficiency results in longer job search times for candidates and higher recruitment costs for employers.
Solution: The Multi-Agent AI Resume Matching System solves traditional job matching inefficiencies by employing four specialized AI agents that work together to provide comprehensive analysis. The Resume Parser Agent extracts structured information from resumes, the Job Description Parser Agent identifies requirements and qualifications, the Matching Agent performs intelligent multi-dimensional scoring using semantic similarity algorithms, and the Report Generator Agent creates actionable insights with personalized recommendations. This system delivers precise percentage matches across skills, experience, education, and soft skills, significantly reducing recruitment time and improving candidate-job alignment accuracy compared to basic keyword-based matching systems.
visualized menu
Team: Selina
Members: Wenhui Tu
Problem: Many people face a problem when trying to order food in unfamiliar restaurants — they cannot understand the menu.
In a diverse city like New York, there are restaurants from all over the world, such as Indian, Korean, Mexican, and Chinese. However, the menus in these restaurants are often in different languages and have no pictures. This makes it very hard for customers, especially tourists or newcomers, to know what to order. They may feel confused, anxious, or even leave without ordering. This issue not only affects the customer experience but also causes restaurants to lose potential business.
Solution: A React app uses the AI to process uploaded menu images, extract structured JSON data, translate dish names into English, and generate visual dish representations. This results in an interactive, user-friendly menu display with images, names, and prices to enhance the dining experience.
Contact
Team: Contact
Members: Bryant Cheng, Issac Zhong, Andy Lin, Kacper Borowski, Evan Haque
Problem: Imagine you're in a situation where you need to contact local authorities, speaking may put you in more danger, reaching for your phone is not an option. Contact provides a secure and silent way to contact and relay information about the dangers to the police. We provide an option of contacting the police so that the situation isn't escalated providing you more time of safety and allowing the police to get to you faster. Where every second counts, Contact makes sure you get as many as possible.
Solution: Transcribes audio, Actively searches cameras and location pings to send to the police. Active live feed from personal camera.
CityHealth Connect
Team: CityHealth
Members: Denelsen Dandi
Problem: Too many uninsured(Zero medicaid) New Yorkers still end up in expensive ERs because booking an affordable clinic visit is slow, paperwork-heavy, and rarely available in their language—while safety-net clinics lose millions each year to empty slots when patients no-show.
Solution: CityHealth Connect’s agent now behaves like a personal “clinician scheduler” that can switch between conversational SMS/WhatsApp and the rich web interface you screenshotted. When a user taps Book Now on any clinic card, the agent pre-loads the clinic’s address, specialty and next-available window, then steers the patient through a three-field form: name, phone, and preferred date. A lightweight rules engine surfaces only open time blocks that match the clinic’s service calendar (9 am–4 pm in your mock) and tags each with service-specific slots—e.g., a 30-minute cardiology block versus a 15-minute general-check-up slot. Behind the scenes, the agent checks Firestore in real time to be sure no one snatched the slot in the last few seconds, then writes the appointment, generates an .ics file, and triggers a Twilio confirmation SMS in the user’s language. Two scheduled functions send reminders at 24 h and 2 h, cutting no-show risk by up to 30 %.
Once the slot is booked the agent bumps the badge on My Appointments (you already show “1”) and renders the visit card with reschedule / cancel options. If the patient later books via SMS—or if a clinic staffer updates status in its EHR—the Firestore listener instantly updates that badge and the list, so the web and text experiences stay in lock-step. The same agent captures referral metadata (specialist, date requested) and pushes status changes (“Approved”, “Scheduled”) to both channels, ensuring patients never lose visibility. Net result: the front-end you demoed feels like a slick consumer app, while the AI agent quietly orchestrates inventory checks, slot locking, calendar generation and bilingual reminders—turning empty chairs into kept visits and giving clinics a live, self-service booking layer without expensive call-centre overhead.
SafeWalk AI
Team: sowcantcode
Members: Soulemane Sow
Problem: SafeWalk addresses the critical safety gap in traditional navigation apps that optimize solely for distance and time. Urban pedestrians, cyclists, and drivers often unknowingly traverse high-crime areas because conventional routing applications lack crime data integration and safety-based route planning. This leaves users vulnerable in unfamiliar locations without actionable safety information or alternative safer path options.
Solution: SafeWalk implements a multi-agent AI system to create safer navigation experiences:
Voice Processing Agent: Converts natural language voice commands into structured route requests, extracting start/end locations, transportation mode, and route priority preferences.
Route Safety Analysis Agent: Processes crime data along potential routes, calculates safety scores, identifies high-risk areas, and generates optimized paths that balance safety and travel time.
Interactive Guidance Agent: Proactively provides safety recommendations, answers user questions about specific routes, and generates contextually relevant safety questions based on the selected route's characteristics.
These agents work in concert to transform raw crime data and geographic information into actionable safety insights, allowing users to make informed decisions when navigating urban environments.