We just launched Meridian on Product Hunt π
It's an automatic worklog that makes sure your work never goes unnoticed. An everyday tool that remembers what you did, so you don't have to.
Support us on Product Hunt: https://t.co/WM8RGo7pVn
Finished #385 out of 1,983 in HackerRank Orchestrate π
Built and deployed an AI agent and learned a lot about agentic systems along the way. On to building bigger things.
@hackerrank#Orchestrate#AI#AIAgents#Hackathon
I don't want to build a better chatbot. I want to build a world model for finance. LLMs read the web, GNNs turn companies into embeddings, and the model predicts how they'll evolve. Inspired by Yann LeCun. #SarvamBuildInHours#AI#WorldModels
Building an AI trading agent using a jepa model that learns from every trade.
MCP gives it tools. RAG + vector memory help it learn from past decisions.
Observe β Trade β Learn β Repeat.
@Paytm@SarvamAI@HackCulture#BuildForIndia
I don't want to build a better chatbot. I want to build a world model for finance. LLMs read the web, GNNs turn companies into embeddings, and the model predicts how they'll evolve. Inspired by Yann LeCun. #SarvamBuildInHours#AI#WorldModels
@odysseyml How tf did I end up here π
VLMs β Physical Intelligence β Yann LeCun β JEPA β World Models β Odyssey.
I was just trying to learn one thing.
Now I'm watching AI interactive worlds and my brain is melting.
Every rabbit hole in AI somehow leads to an even bigger hole.ππ
Inspired by Bayesian thinking (Roger Labbe's Kalman filter notebooks), I'm exploring GraphSAGE to learn which cone detections are valid instead of simply discarding noisy ones. Also looking at using large orange cones as landmarks for loop closure.
SLAM stack for Formula Bharat Driverless car.
Replacing TCP with shared memory (mmap) for IPC drastically reduced latency and improved system responsiveness.
A good reminder that architecture often matters more than algorithmic optimizations.
#SLAM#EmbeddedSystems#Linux#Rust