"NATO Hackathon Winner: AI-Powered Drug Search", Taras Kloba
•
0 likes•110 views
This is a session that details how PostgreSQL's features and Azure AI Services can be effectively used to significantly enhance the search functionality in any application. In this session, we'll share insights on how we used PostgreSQL to facilitate precise searches across multiple fields in our mobile application. The techniques include using LIKE and ILIKE operators and integrating a trigram-based search to handle potential misspellings, thereby increasing the search accuracy. We'll also discuss how the azure_ai extension on PostgreSQL databases in Azure and Azure AI Services were utilized to create vectors from user input, a feature beneficial when users wish to find specific items based on text prompts. While our application's case study involves a drug search, the techniques and principles shared in this session can be adapted to improve search functionality in a wide range of applications. Join us to learn how PostgreSQL and Azure AI can be harnessed to enhance your application's search capability.
Recommended
More Related Content
Similar to "NATO Hackathon Winner: AI-Powered Drug Search", Taras Kloba
Similar to "NATO Hackathon Winner: AI-Powered Drug Search", Taras Kloba (20)
More from Fwdays
More from Fwdays (20)
Recently uploaded
Recently uploaded (20)
"NATO Hackathon Winner: AI-Powered Drug Search", Taras Kloba
- 1. ADVANCING DRUG SEARCH WITH POSTGRESQL AND AZURE AI Taras Kloba Senior Solutions Architect, Data & AI at Microsoft POSETTE - 2024
- 2. PANACEA Taras Kloba Senior Solutions Architect, Data & AI at Microsoft • Founded a volunteering IT group for Ukrainian army systems, winner of the TIDE NATO Hackathon and Ukraine Defence Hackathon. • Co-leader of PostgreSQL Ukraine and Big Data Community • More than 14 years of experience in Data Engineering • Ph.D. in Economics • Winner of the Ukrainian IT Awards 2019 in Software Architecture • Certified Cloud Architect & Data Engineer on Google Cloud, Microsoft, and Amazon Web Services. Microsoft Certified Trainer • Father of three daughters
- 3. PANACEA the challenge. understanding our challenge 3
- 4. PANACEA Medication Match-Up. When a new war starts, NATO troops from different countries unite, each with their medical supplies. Our challenge was to make an app that helps find the same medicine across different country brands. 4 The Suwalki Gap: A Hypothetical Flashpoint PANACEA
- 5. PANACEA the solution. innovation born of necessity 5
- 6. PANACEA Inna Olkhova. The diagram shows our app's ability to match different medicine brands through their active ingredient, Acetaminophen. Inna’s idea helps find the same medicine, whether it's Tylenol in the US, Panadol in the UK, Paracetamol in the EU, or Dafalgan in France. known as the "Bee Paramedic" 6
- 7. PANACEA Tata Kepler. From Artistry to Aid: Shaping Accessible Care in the Field 7 Tata showed us how medicine sorting works, especially when volunteers might not know medical terms. She introduced us to using the ATC system in our app, making it easier for everyone to find and organize medicines correctly.
- 8. PANACEA Daryna Smolnikova. Daryna shared the idea of using NFC, QR codes, and short links to store and share information about the wounded. She emphasized that sometimes, sharing this information, even with the enemy, is crucial to saving lives. Immediate Action: Key to Battlefield Survival 8
- 9. PANACEA Iryna Soloshenko. Train Evacuations in Ukraine: Pioneering Lifesaving Journeys 9
- 10. PANACEA tech specs. engineering precision for lifesaving solutions 10
- 11. PANACEA 11 Architecture. Azure Database for PostgreSQL with the Azure_AI extension
- 12. PANACEA 12 Architecture. Azure Database for PostgreSQL with the Azure_AI extension Can be replaced with azure_local_ai extension Can be replaced with Data API Builder
- 13. PANACEA 13 ILIKE. The ILIKE operator in PostgreSQL is used for case-insensitive pattern matching. It functions similarly to the LIKE operator but ignores the case of the character SELECT * FROM drugs WHERE drug_name ILIKE 'tylen%'; SELECT * FROM drugs WHERE drug_name ILIKE '%tylen%';
- 14. PANACEA 14 TRIGRAMS. A trigram is a group of three consecutive characters in a string. By comparing these trigrams, PostgreSQL can measure the similarity between strings. Useful for finding similar strings despite minor spelling errors and variations. CREATE EXTENSION pg_trgm; SELECT drug_name FROM drugs WHERE drug_name % 'tylelon' ORDER BY similarity(drug_name, 'tylelon') DESC;
- 15. PANACEA 15 VECTOR SEARCH. Vector search uses vector representations of text to find similar items. By comparing these vectors, PostgreSQL can measure the similarity between text descriptions. CREATE EXTENSION azure_ai; ALTER TABLE drugs ADD COLUMN vector VECTOR GENERATED ALWAYS AS (azure_ai.vectorize(drug_name)) STORED; SELECT drug_name FROM drugs ORDER BY vector <=> azure_ai.vectorize('pill for pain management') LIMIT 5;
- 16. PANACEA 16 VECTOR SEARCH. Vector search uses vector representations of text to find similar items. By comparing these vectors, PostgreSQL can measure the similarity between text descriptions. CREATE EXTENSION azure_ai; ALTER TABLE drugs ADD COLUMN vector VECTOR GENERATED ALWAYS AS (azure_ai.vectorize(drug_name)) STORED; SELECT drug_name FROM drugs ORDER BY vector <=> azure_ai.vectorize('red pill for headaches') LIMIT 5;
- 17. PANACEA 17 DRUGGPT (AI RAG ASSSISTANT). RAG (Retrieval-Augmented Generation) is a hybrid approach that combines retrieval and generation techniques to enhance the quality of AI responses. It first retrieves relevant information from a knowledge base or database and then uses this information to generate more accurate and contextually relevant answers.
- 18. PANACEA 18 Search. Layer 1 Layer 2 Layer 3 Layer 4 Layer 5 Layer 6 Layer 7 Machine learning models for semantic representation, ranking for the rest of results. Exact and pattern text search (ilike%), trigram similarity calculation, and ranking for best results. Similarity search with threshold 0.7, ranking for best results. Machine learning models for semantic representation, with a threshold of 0.9, ranking for best results. Pattern text search (%ilike%), trigram similarity calculation, ranking for best results. Full-text search in summary, description, and indication columns. Similarity search with threshold 0.5, ranking for best results. Yes No Is result found? Yes No Is result found? Yes No Is result found? Yes No Is result found? Yes No Is result found? Yes Is result found? No
- 19. PANACEA 19 OCR Search. [Tylenol], [Acetaminophen], [Pain], [Reliever], [Fever], [Reducer], [Extra], [Strength], [FOR], [ADULTS], [500], [mg], [each] Database
- 20. PANACEA 20 Knowledge Graph. Powered by DrugBank.com database
- 21. PANACEA demo. 21
- 23. PANACEA get in touch. Taras Kloba aka.ms/taras 23