Faruk Ajibade

her to you 😂😂

man doesn’t want stress😂😂

shipped hybrid search for knowlia 🔍 vector (semantic) + bm25 (keyword) = better retrieval why? vector search struggles with: • product names • technical jargon • error codes hybrid search fixes this github.com/1FarukDev/knowlia ⭐ #ai #rag #buildinpublic

you might want to check the link

monday thoughts on learning ai engineering 🧠 been building knowlia (rag engine) from scratch this past week the real lesson isn't about the project it's about the learning process here's what changed my perspective 👇 #buildinpublic #ai

morning run….. stay fit 🫡

added reranking to improve answer quality the problem: vector search finds "similar" content but not always the most "relevant" the fix: retrieve 10 → rerank → keep best 5 here's the difference 👇 #ai #rag #buildinpublic

rag explained in code 👇 the query pipeline: • retrieve relevant chunks (vector search) • generate answer with llm (gpt-4) • return grounded response full version on github with error handling, logging, validation github.com/1FarukDev/knowlia #ai #rag #buildinpublic

stay fit ✌️

“egbonnn”

data science student by day. software engineer by night. ml model trainer in between. i build ai-powered apps and ship real client work — all while figuring out data science. follow me to walk through the journey. 🔗 github.com/1farukdev 💼 linkedin.com/in/farukajibade