“AI is just not a buzzword anymore. It’s about which group is implementing it sooner, and at scale,” stated Kanika Gupta, Options Architect at Redis, as she opened her deep-dive session at DevSparks Hyderabad 2025. For Kanika, the race is not about whether or not corporations will undertake AI, however how rapidly they will overcome its bottlenecks of pace, accuracy, and price.
She reminded the viewers that almost all builders know Redis as a caching answer, however that story is altering. “Redis is past caching,” she stated. “We’re the working reminiscence layer for AI.”
Kanika defined that Redis is positioning itself as a spine for next-generation purposes from conversational brokers and suggestion techniques, to real-time analytics and fraud detection.
Tackling latency, value, accuracy
One of many largest challenges, Kanika famous, is latency.
“It generally takes 5 to 10 seconds to generate an LLM reply; we don’t have that sort of time. Pace is one thing we can not compromise on,” she defined.
The price of repeatedly hitting giant language fashions, mixed with the chance of hallucinations and the complexity of integrating hundreds of latest AI instruments, solely compounds the issue.
Her session centered on how Redis is tackling these points head-on. By serving as a vector database for retrieval-augmented era (RAG), Redis can retailer embeddings and metadata collectively, enabling hybrid and full-text search at in-memory speeds. “Redis Enterprise serves as a completely featured vector database, requiring no further setup or set up,” Kanika stated, declaring that Redis has been benchmarked as 62% sooner than the second-best vector database.
Past retrieval: Semantic caching and agentic reminiscence
Kanika additionally highlighted Redis’s function in fixing certainly one of AI’s costliest inefficiencies: redundant queries to LLMs. “Anytime I ask my cost app the identical query another person has requested, it shouldn’t should hit the LLM once more,” she defined. The answer is semantic caching, which Redis now gives as a service by means of LangCache, permitting builders to cache and retrieve responses for semantically comparable queries.
The opposite breakthrough she described was Redis’s potential as a reminiscence layer for brokers. “If I ask a bot within the morning to ebook a flight, and within the night I ask it to vary that flight, it ought to have the context,” Kanika stated. “That’s the place Redis comes into the image…because the reminiscence of the agent.”
Past caching and retrieval, Redis is broadening its function in knowledge infrastructure. With Redis Information Integration, builders can maintain supply techniques and caches in sync in actual time, whereas RedisFlex allows tiered storage, maintaining scorching knowledge in reminiscence and shifting heat or chilly knowledge to SSDs. Redis can be investing in ecosystem readiness, providing consumer libraries in a number of languages and integrations with standard AI frameworks similar to OpenAI, LangChain, and AWS Bedrock.
Efficiency: The non-negotiable
A lot of Kanika’s session circled again to a single precedence — delivering efficiency at scale. Whether or not powering chatbots, fraud detection, or suggestion engines, Redis is positioning itself not as a supporting act however as a crucial layer within the AI stack. “It’s an in-memory vector DB plus accuracy, so pace and accuracy could be achieved collectively,” she stated.
Her closing remarks captured the ambition Redis now carries into the AI period: “Redis is not only one caching answer. Begin utilizing it as your reminiscence and intelligence layer for AI.”
Elevate your perspective with NextTech Information, the place innovation meets perception.
Uncover the most recent breakthroughs, get unique updates, and join with a world community of future-focused thinkers.
Unlock tomorrow’s traits at present: learn extra, subscribe to our e-newsletter, and develop into a part of the NextTech group at NextTech-news.com
