AC215 Project Showcase
(December 9th 2024 - 9:00 AM - 11:00 AM)
| Description | |
|---|---|
| 1. DecAide Our tool allows for celebrity stylists, people who need an efficient way to understand historical fashion references, to style their clients using clothing featured on high fashion runways. Our app takes images of outfits, identifies the year that each outfit most closely resembles, and provides recommendations and information to the user based off of that decade and its popular trends. |  |
| 2. SnapChef An app that identifies dishes from images, generates detailed recipes, and tailors them to individual preferences and dietary needs could empower food enthusiasts to recreate dishes seen on social media or in restaurants effortlessly. |  |
| 3. CareConverse CareConverse is an LLM-powered chatbot to surface personalized information on optimal chronic disease management. By utilizing a large language model (LLM) combined with Retrieval-Augmented Generation (RAG), CareConverse will deliver concise, pertinent context to patients looking for clarity, reassurance and direction on their disease experience, along with citations and links to relevant content. This approach ensures transparency and trust, as users can verify the information and further explore the literature. |  |
| 4. Pourfect AI PourfectAI is a conversational chatbot bartender that blends expert mixology knowledge with the convenience of a personal recommendation system. PourfectAI suggests the perfect drink based on user preferences, mood, ingredients on hand, or even dietary needs—whether it’s a fresh mocktail or a classic cocktail with a twist. PourfectAI – where every pour is perfectly yours! |  |
| 5. ClinIQ To develop a diagnostic medical assistant tool where clinicians can input symptoms/patient notes into a chatbot and receive possible diagnoses and recommended medications to guide faster, more accurate diagnoses. |  |
| 6. EmpathOS Our project, EmpathOS, is a mental health chatbot providing immediate, confidential, and expert support for individuals facing mental health challenges. Leveraging fine-tuning and Retrieval-Augmented Generation (RAG) on Google Cloud, it delivers 24/7, evidence-based guidance, specifically tailored for groups such as teenagers and college students, ensuring scientifically grounded, context-aware responses. |  |
| 7. 1nvestSense – LLM-powered stock advisor for GenZ investors 1nvestSense is an AI-powered application that leverages LLMs to deliver real-time, actionable insights from market news, tailored specifically for retail investors, particularly Gen Z. |  |
| 8. CrimsonChat Computer Science graduate students enter Harvard expecting to be productive and have effortless access to resources, but they hit the obstacle of bureaucratic overhead. CrimsonChat addresses the problem through a friendly, 24/7 available assistant with comprehensive knowledge of departmental resources. For stressed students, CrimsonChat also offers 24/7 mental health support. |  |
| 9. MoodSync Introducing a cutting-edge music recommendation tool designed to understand and enhance your listening experience. Simply share how you’re feeling and your favorite music elements, and our tool leverages AI to craft playlists that perfectly match your current emotional and musical appetite. |  |
| 10. Priva-see Priva-see is your path to data privacy for any terms and conditions. Simply upload a set of terms and conditions and get a privacy grade along with a summary of key issues. Or input your privacy preferences and receive personalized app recommendations that meet your privacy needs. |  |
| 11. SmartEats SmartEats is an innovative AI-powered application that analyzes food images, predicts disease risks, and provides personalized nutrition advice. By simply uploading a photo of their meal, users can get an estimate of nutritional components, calorie count, and tailored dietary suggestions. While designed for a broad audience, SmartEats is especially valuable for individuals with specific dietary needs - whether they’re looking to lose weight, track daily calories, optimize nutrition for fitness, or manage diet-related health risks. |  |
| 12. Protein_Server This project aims to develop a user-friendly web application for researchers in protein research. An interactive chatbot will be built to answer general questions on protein structures, functions, and characteristics. A protein LM based tool will allow users to input protein sequences and receive protein function prediction, such as thermostability. |  |
| 13. Capy Running Capy Running gives you full-control over your training without the need to get yet another app. We are an on-demand 24/7 AI-powered personal running coach that works by 1/10 of the cost of a normal coach and all through Whatsapp Messaging. With Capy Running, you get a personalized training schedule, reminders, performance recording and much more. |  |
| 14. ByteBites We designed and developed an AI-powered recipe web app that processes shopping receipts to automatically detect purchased food items and suggest corresponding recipes. Users can also specify dietary preferences, such as vegan or gluten-free options. Additionally, we plan to integrate a nutrition tracking feature to enhance the functionality of our app. |  |
| 15. SpiritMatch In general, getting started with cocktail making at home can be quite intimidating. There are a lot of things to keep in mind (such as drink ingredients, tools, garnishes, whether or not to add ice, etc.). Therefore, the aim of this project is to develop an application that can allow for users to rely on a virtual bartender, Eustace, who will easily generate cocktail recipe recommendations based on the user’s mood, situation, and/or household ingredients. With every interaction, users can expect a unique and thoughtful experience, where cocktail-making becomes both a creative journey and a delightful discovery, tailored just for them. |  |
| 16. Yarn Master It’s really common to find crochet products we’d love to make ourselves, but finding clear, step-by-step instructions can be really challenging. Currently, there isn’t an AI-powered tool that can effectively generate meaningful crochet instructions. We came up with the idea of developing a deep learning model specifically designed for crochet products. Users would be able to upload pictures of crochet items they find interesting, and the model would generate detailed instructions for recreating those products. |  |
| 17. Bloodwise Bloodwise provides users with easy explanations of lab test results based on the provided results and a symptom summary. The application suggests what the results could mean, and if abnormal, suggests lifestyle changes, specifying to consult with a physician as well. |  |
| 18. Billionaire.ai Billionaire.ai is a professional, real-time financial advisor powered by advanced LLMs, delivering concise insights and rich visualizations tailored for investors of all levels. Transform your investment strategy with trusted data and expert recommendations in an instant. |  |
| 19. Sales Mate We built an LLM/RAG-enabled customer sales rep training tool. Our product provides two main functions for sales training teams: a customer simulator that enables sales reps to practice initiating and closing sales conversations and a copilot knowledge helper that provides conversation advice in real-time. |  |
| 20. Daily Meal Assistant Our project develop an app that serves as your personal meal assistant. This app recommends recipes based on the ingredients you have on hand, which you can simply upload via images. By combining Large Language Models (LLMs) with Retrieval-Augmented Generation (RAG) and dynamic API integrations, our app delivers insightful suggestions for both home cooking and dining experiences. Transform the way you eat with our smart, responsive meal assistant at your fingertips. |  |
| 21. VeritasTrial This project aims to develop an AI-powered system that enables users to efficiently search and explore clinical trials within a database. The system will utilize a fine-tuned language model (LLM) to identify and retrieve the most relevant trials based on user queries. It will then allow users to interact with specific trials and provide accurate responses by leveraging both the fine-tuned model and structured clinical trial data stored in the database. |  |
| 22. Tripee Tripee is an LLM-powered trip itinerary generator that optimizes routes and incorporates map visualization. By leveraging state-of-the-art fine-tuning and Retrieval-Augmented Generation, our solution generates thematic itineraries tailored to users’ specific needs, including city, duration, and type of trip, without the need for hosting large travel datasets. |  |
| 23. Harvard Academia Atlas The Harvard Academia Atlas helps students navigate their academic journey by offering personalized course recommendations based on their backgrounds, interests, and goals. It suggests courses, explains their relevance to academic and career paths and helps create conflict-free schedules for easy management. |  |
| 24. ECGuide Interpreting electrocardiograms (ECGs) is a critical skill for healthcare providers, yet ECG education has remained largely unchanged for over 50 years. ECGuide revolutionizes this learning experience by offering personalized, interactive ECG education. Users can upload an ECG, and our platform guides them through a detailed analysis, highlighting key diagnostic features and helping them build a deeper, more practical understanding of ECG interpretation. |  |
| 25. Adversarial Playground Adversarial Playground is an app that enables users to investigate the consequences of adversarial attacks on image classification problems. Users are able to interact with the Adversarial Robustness Toolbox to perform preset attacks against models to demonstrate their efficacy while having the option to observe how these attacks can be mitigated. |  |
| 26. AI4Good A large portion of any research is limited to academic journals and does not result in practical societal changes. The goal of this project is to close that divide by developing a platform that links AI research groups with organizations and associations that could gain from their research. With the help of Large Language Models (LLMs), we aim to discover academic papers that have the capacity to tackle societal issues, promoting valuable partnerships. |  |
| 27. PrepPal PrepPal is an LLM-powered and sustainability-focused recipe assistant that removes the annoying aspects of cooking a meal: all the planning! Based on the user’s preferences and cooking time, PrepPal suggests exciting recipes that prioritize the usage of already available ingredients, thereby allowing the user to discover new meals effortlessly while reducing food waste at the same time! |  |
| 28. Innit Innit is an immersive language learning app that focuses on giving users more practical language skills that are applicable to their daily lives. We track users’ language levels and we will automatically recommend them media every single day at their specific language level which they can interact with via features such as automated summarization and Q&As. |  |
| 29. PillRx PillRx is a Computer Vision app for detecting pills identity from images of pills, followed by a warning about potentially dangerous interaction with alcohol and the option to chat with a RAG-based LLM to learn more about the identified pill. |  |
| 30. Fashion AI Fashion AI is an AI-powered fashion search platform, allowing users to conduct natural language-based search and styling recommendations without endless browsing. |  |
| 31. gAIn gAIn is you automated health and fitness coach, providing expert opinions and suggestions about your health and fitness journey based on up-to-date research and your detailed fitness and activity history. We leverage a state-of-the art LLM for chat interactions with the coach, fine-tuned on health and fitness data, and supplemented with data on the user’s historical activity through RAG on data such as Strava history. |  |
| 32. Burning Bush BurningBush is an AI-driven bushfire prediction and mitigation platform designed to safeguard ecosystems, property, and lives. By combining historical geospatial data and multispectral imagery from Sentinel-2, the system predicts bushfire risks in specific regions. Users can search for their property and assess fire probabilities that will allow them to develop timely mitigation plans. BurningBush empowers individuals and first responders to take proactive measures, reducing the devastating impacts of bushfires on vulnerable areas. |  |
| 33. Reverie Reverie is an AI-powered knowledge exploration platform that redefines learning through intuitive navigation and dynamic pathways |  |
| 34. Game, Set, Match Game, Set, Match is an online application that aims to predict the winner between two profession tennis opponents. The prediction model is trained on a large public data set of past matches and their results. |  |
| 35. CarFever In this project, we aim to develop an application that can accurately identify the car model, make, and year from user-uploaded photos. |  |
| 36. MediScribe MediScribe allows physicians to record conversations with their patients and generate an initial draft of an electronic medical record, which they can later review and edit. This essentially automates a time-intensive manual process and allows physicians to dedicate more time towards providing healthcare to their patients rather than recordkeeping. |  |
| 37. Paper with Data Paper with Data is a search engine enabling scholars to search for papers with high relevance with the exact data variables they inputed. It aims to provide a fine-grained search logic to accelerate the research process. |  |