Built EcoTwin for @WeMakeDevs' Environment Hacks! Modeled Zürich using satellite, weather and air quality data. Remove or change city elements to simulate environmental impacts, quantify tradeoffs and get (optional) LLM explanations. https://t.co/0mWzryEyXi #WeMakeDevs#Hackathon
Amazon's ML Challenge is nearing its end rn, and our first day score (ai slop code which wasn't even submitted on first day✌️😭) was some 0.43. Ramped it up to 0.56 (not much of an improvement but still had to spend hours tweaking params),now we're hoping we atleast touch 70😭🥀
Just finished the Agent Harness hackathon by WeMakeDevs. We (myself, @AdarshJha2709, @blxnk_01, @ChiragWattamwar) created an ai agent that automated safe SDK migrations (tested using stripe API and the official migration guide).
repo:https://t.co/7X4sgutHVI
#WeMakeDevs
@datachad Yeah I skipped it too, but there wasn't any point anyways the model was giving the results that were too far off from the intended schema design. It was skipping sections and fabricating data so the results never reached the formatter for file generation.
I finetuned qwen3-4b model for generating content for university assignments using previous files. Not exactly my first time finetuning, still had to use a lot of LLM assistance. Its still incomplete, mostly due to lack of data. Suggestions are welcomed!
https://t.co/M9RZvnZ7R3
Built a RAG application that uses a local model (llama 3.2:3b) for retrieval and generation, mainly focused on confidentiality. Right now, it still has many limitations. Hoping to improve this moving forward.
Feel free to drop suggestions and PRs!
https://t.co/mzHq3byjb2
Generated an automation script for google apps script to automate creation of a DBMS google docs assignment for university. Only upload named screenshots in a drive folder and run the script, and the doc is generated! Should've started using this a LOT sooner.