Roro

Building LLM applications is a different dimension of software engineering altogether. Having cut through the fluff around generative ai and worked on several real LLM applications, here are few things that we do at Roro and severely enjoy. ?? Eliminate. (if the task shouldn't require LLM) ?? Categorise & Strategise. ?? Unstructured data science. ?? Development of LLM applications using LangChain. ?? RAG systems using vector databases like Pinecone, FAISS and ChromaDB. ?? Simple Agents based on ReAct prompt. ?? Complex Agents and multi-agent systems using LangGraph and flow engineering. ?? Monitor and optimise using LangSmith. ——————————————— Different LLMs have different behaviour and target use-cases. Here are a few things we do to optimise (cost+performance) the use of LLMs. ?? Different models for different use-cases, even within the same project. ?? Low cost or local models (using Ollama) for simpler tasks. ?? Decide if a use case require an LLM to be fine-tuned. ?? Pandas and Seaborn to prepare dataset for fine-tuning. ?? Fine-tune LLMs using platforms like OpenAI, LlamaFactory, Unsloth, Gradient and HuggingFace. ?? Case-optimise latency-accuracy trade-off during fine-tuning through quantisation, LoRA and pruning. ——————————————— It’s a very exciting time to jump on the Generative AI bandwagon. If you have a thought or use case around your business, we’ll love to talk and brainstorm. Visit our updated website at?www.roro.io?and a get a quote for your Gen AI project. #llm #generativeai #ai #langchain #vectordatabase #rag

Working with?Skin Dossier?has been an incredible journey. Their AI-powered diagnostics are leading the way consumers approach skincare, haircare, supplements, and more—bringing precision and personalization to the forefront. A big thank you to our visionary partner?Sindhya Valloppillil?for her kind words and for trusting Roro to deliver high-quality, impactful results.?Here's to creating smarter, impactful solutions together! #ClientTestimonials

Remember, you don’t need every detail of your product figured out— we’re here to brainstorm with you and introduce ideas you might not have considered. Book a meeting with us and let's get cracking - https://buff.ly/4dvHxE7

We've loved working with Eyrus, the worksite intelligence platform that brings automation and smart insights to construction teams worldwide—from onboarding to access control and more. Our goal was understanding Eyrus’ unique needs and deliver a solution that would truly support their mission. Thanks to Ben Brautigam and the entire Eyrus team for the trust and partnership. Special shoutout to Ben for the thoughtful words. #ClientTestimonial

Open AI just released a framework to do lightweight multi-agent orchestration called Swarm. They call it ‘educational’ but after going through their GitHub page, it seems it will instantly kill Crew.ai. Swarm implements flow engineering at a ‘lighter’ scale, more autonomously, while LangGraph does it extensively and with more control on the flow. Here is a gist of what Swarm does: - Route (triage agent) a situation between available specialised agents. - Maintain context variables (really like this!) gathered over the conversation. Its similar to graph state in LangGraph, just more at the disposal of the LLM. - Execute tools, which can wrap any injectable source like an API call or Vector DB retrieval. - Handoff so that the triage agent can assign it to a different agent or finish the process. ————————————————— It’s too early to start using Swarm in real projects. Currently, it’s more of a way for us to validate and align our thoughts with OpenAI engineers about multi agent situation. The only thing I employ in my LLM Agent projects and do not see here or in LangGraph is the formal definition of ‘Actions’. I don’t think every situation requires multi-agent. A lot of them are more of a situation with one agent that performs several actions to accomplish a task. Thoughts? #openai #agents #llm #ai

#MeetTeamRoro Meet Aitreya Verma, a Frontend Software Engineer who skillfully navigates between React and various other technologies to ensure project success. Swipe the post below to know more about Aitreya. #TeamRoro

As 2024 wraps up, we’re taking a look at some of the top programming languages of the year, in this blog post written by Anuj Singh. Whether you're just starting out or a seasoned developer, there’s something here for everyone. Check it out - https://buff.ly/3AgPYFV #ProgrammingLanguages #JavaScript #Python #Golang #Ruby #Kotlin

We're thrilled to have partnered with HumanFab, a leader in advancing human/product/technology interactions to enhance customer experiences for manufacturers. Huge thanks to Thomas Mousques for the kind words and for trusting us with your vision! #ClientTestimonial

A bad UI can break a brand fast, but a well-designed one can boost conversion rates by up to 200%, with good UX increasing that to 400% (Forrester). Intuitive design builds trust and drives revenue. At Roro, we prioritize design because it shapes your customer's first impression. Book a meeting with us to explore how we can make your product's design more intuitive - https://buff.ly/4dvHxE7 #IntuitiveUI #DesignFirst #RoroSolutions

#MeetTeamRoro Meet Shamsul Zama, our Senior Software Engineer specializing in Shopify and Ruby on Rails who just hit his 2-year milestone with us! ?? When he's not solving complex problems, you'll find him soaking up nature, listening to stories, or finding peace in prayer after a busy day. Swipe the post below to get to know Shamsul better. #TeamRoro