I'm Ratish! Welcome to my Github. I specialize in full stack and applied AI/ML software engineering and development. In the past, I've done freelance work building bots for NFT/Web3 orgs, built some personal projects, and contributed open source code as part of GSoC 2024.
Tech Stack:
- Languages:
,
- Frameworks:
,
,
,
,
,
,
Open Source Contributions:
- Google Summer of Code 2024: I worked on Arviz - a Python package and project for exploratory analysis of Bayesian (probabilistic programming) models. More details on this blog post
Selected Works/Projects:
-
Crustdata API Support Bot: Next.js + Vercel AI SDK based AI chat support agent. Ask any questions about Crustdata's API. Uses RAG (with Crustdata API docs) to answer questions. Also has an authentication system and stores past chat histories. Repository private for now.
-
BOPE-GPT: Next.js+ FastAPI based web app to orchestrate and perform the BOPE process (Bayesian Optimization with Preference Exploration) for multi objective optimization tasks. A prompted large language model is used as a preference selector and pareto front + gaussian process visualizations can be viewed. Also uses MongoDB, Pytorch, Cohere API, and Plotly. Repository
Based on a technique introduced in a 2022 paper by researchers at Meta and Cornell University and some work done automating one of its steps with an LLM in a 2024 hackathon I was part of.
-
Poiesis Ex Machina: Astro + React + Svelte serverless AI poetry generator with cyberpunk/tech-noir aesthetic. Enter a prompt and pick an archetype (Eristics, Jungian, Dungeons and Dragons). Uses Together.ai for image and text-to-image generation. Repository
-
Web3 + Custom Bots: Super-repo/Portfolio of old bots I made for freelance clients. Built with Python and Javascript and several 3rd party APIs and API wrappers, for Discord, Twitter, Telegram.
Currently Learning More About:
- AI-augmented coding patterns and methodology- including working effectively with AI agents (productivity enhancements here are potentially so massive, and it likely will become the norm in the next few years)
- ML/AI-engineering techniques and best practices- actually using AI/ML for features in fullstack apps, LLM usage techniques like fine-tuning, RAG, prompt-chaining, MoE, etc. Making full use of these kinds of emerging intelligence wherever possible and advantageous.
- Auxiliary Devops-y tech apps need in production (CI/CD, containerization, queues and in memory caches, cloud provider tools for scaling, database sharding, the intricacies of a backend web server- that sort of stuff). These are of course fundamental software engineering concepts, especially for anything pushed to production- necessary to scale products smoothly.