Here Technologies aims to provide a complete, accurate and easy to use digital representation of the physical world. Students on this team will leverage large language models (LLM) and Geographic Information System (GIS) technologies to group related points of interest together for improved navigation.
Abstract:
Here Technologies is a location data and technology company that created the first digital map over 35 years ago. Here provides a wide range of value-added mapped information to provide accurate, easy to use representations of the physical world. In their Places domain, Here has come across situations where a collection of places or points of interest (POIs) could naturally be considered as a single set of related nearby entities.
There are several uses cases for having inter-places relationships.
- Search record prioritization and co-discovery – When a user search for a place by name, the records appearing with the name can be properly organized as a parent-child list.
- Rest areas – drivers can understand what kind of other facilities such as fuel stations, EV charging stations, restaurants available at a given rest-area.
- Visualization – related places can be color coded (e.g. all the related university facilities can be represented similar manner).
- Trip Planning – related places can be used for suggestions in trip planning e.g.
- Museum campuses.
- Parking facilities around large venues such as stadiums.
- EV charging stations.
- Places with multiple access points – large venues can have more than one access point represented as individual places having different routing (E.g. Chicago Union station). End users will have options to pick based on their preference.
One good example of this is the buildings, infrastructure, and resources around university campuses. The student team will add to the Here technology base by creating a proof of concept (POC) function, demonstrating automatically derived relationships between separate location elements on Michigan’s campus, and grouping them together under a single umbrella. We anticipate that the team will utilize some/most of the following techniques:
- Use of other available map data. e.g. cartographic layer, road network to determine proximity/accessibility in addition to simple position (latitude, longitude)
- Use of place categorization to identify places that can be grouped based on categories
- User of LLM models via prompt engineering to identify potential records that can be considered related
Note: While the initial POC is related to universities, we expect the solution to be applicable to other scenarios such as shopping malls, museum campuses, government facilities (police, fire, public services, etc.)
Elements that might be incorporated into the group definition for college or University include:
- Department Buildings
- Libraries
- Dormitories
- Sports Facilities (Stadiums, Gymnasiums, Swimming Pools, etc.)
- Administrative Facilities
- Car Parks
- Transport Access Points (bus/train stops, bike rentals, etc.)
It might also include items that are not adjacent, or are less closely related:
- Co-op housing
- Fraternities and sororities
- Relevant hotels or restaurants
- Parks/botanical gardens
- Hospitals and medical centers
Impact:
Knowing relationship information of a given place provides the opportunity to offer additional capabilities to the end-user, such as: understanding parking options, locating EV charging, and visit planning (park at one place and walk/shuttle/bike etc.).
Scope:
Minimum Viable Product Deliverable (Minimum level of success)
- Complete a literature review of all relevant materials that are publicly available and/or knowledge from the sponsoring organization
- Identify a range of solution ideas and investigate them with mini projects to evaluate them for priority/feasibility. These might include:
- Use of name similarities (including partial names)
- LLM to determine contextual relationships (e.g. while “East Quad dining hall” has no reference to U-M, it is understood as a facility related to the university, by contextual knowledge)
- Use of distance filtering along with other signals (e.g. Just because Joe’s Pizza is next to Department of Mathematics, we don’t consider those two places related)
- Use of building footprints, campus cartography
- Use of road network (traverse-ability)
- Complete an initial proof of concept application based on the most successful ideas
Expected Final Deliverable (Expected level of success)
- Demonstrate the proof of concept over a range of use cases, and determine adjustments that need to be made to address shortcomings.
- Complete a more refined, second version of the proof of concept – a generalized solution that can be applied to similar establishments (e.g., shopping malls, plazas, resorts, museum campuses, healthcare facilities, government buildings, rest areas, etc.)
Stretch Goal Opportunities: (High level of success)
- Automatically curating a list of points of interest from elements from within the University group
Below are the skills needed for this project. Students with the following relevant skills and interest, regardless of major, are encouraged to apply! This is a team based multidisciplinary project. Students on the team are not expected to have experience in all areas, but should be willing to learn and will be asked to perform a breadth of tasks throughout the two semester project.
General Programming (2-3 Students)
Specific Skills: General interest and skills in programming
This project will use Java, Scala or Python
Completion of EECS 281 (or equivalent) is required.
Likely Majors: CS, DATA
Machine Learning (2-3 Students)
Specific Skills: Understanding of LLM, RAG, and related ML concepts.
Understanding of Cloud Native development. Full-stack development experience is a plus.
Completion of EECS 281 (or equivalent) is required.
Completion of EECS 445 or other ML classes would be a plus.
Likely Majors: CS, DATA
GIS domain experience (1-2 Students)
Specific Skills: Practical experience working with geographic datasets and cartography
Completion of EAS531/Environ411 is a plus
Likely Majors: CS, SEAS
Additional Desired Skills/Knowledge/Experience
If you have any of these characteristics, highlight them on your Experience and Interest Form and talk about them in your (optional) one way video interview.
- If you have prior experience in any of the chosen tech stack, please indicate this in your personal statement:
- Languages: Java, Scala or Python
- Platform: Cloud Native (AWS preferred)
- Libraries/SDKs: Open source
- Data: HERE Technologies will provide necessary data sets via S3 buckets or zip files
- Interest or experience in geographic information systems and mapping
Sponsor Mentor
Buddika Gajapala
Buddika is a Senior Principal Engineer at HERE Technologies with 25+ years of experience in Software Engineering in various domains, including Consumer Banking, Trading, Mortgage, and Geospatial/Maps. He has broad and deep experience in designing and implementing large scale, high-performant systems, and Cloud computing. Buddika Gajapala has two patents.
Sponsor Mentor
Sugih Jamin
Associate Professor, Computer Science Engineering
Weekly Meetings: During the winter 2025 semester, the Here Technologies team will meet on Tuesdays from 12:30 – 2:30 PM. Location TBD.
Work Location: Most of the work will take place on campus in Ann Arbor.
Course Substitutions: CE MDE, ChE Elective, CS Capstone/MDE, Data Science MDE/Capstone, EE MDE, CoE Honors, SI Elective/Cognate
Citizenship Requirements: This project is open to all students. Note: International students on an F-1 visa will be required to declare part time CPT during Winter 2025 and Fall 2025 terms.
IP/NDA: Students will sign IP/NDA documents that are unique to Here Technologies.
Summer Project Activities: No summer activity will take place on the project.
Learn more about the expectations for this type of MDP project
Sugih is an Associate Professor of computer science at the University of Michigan and possesses more than 30 years of experience in Internet measurement, protocol and infrastructure design and deployment, computer graphics, and iOS and Android native development. Sugih is an experienced MDP mentor.