Allwins Rajaiah

06

Projects

Personal experiments, labeled as such.

Personal and experimental work. These are not production engagements and are not client projects. Repositories, demos, and lessons are unpublished, so those fields stay empty.

Personal / experimental

Local SLM

Problem: A personal experiment with a smaller language model running locally, so prompts stay off a hosted API.
Architecture: A Llama-based local model behind a thin application boundary. Retrieval, tools, and an evaluation loop are not specified.
Technology: Llama-based local model. TODO: runtime, hardware, and serving stack are not documented.
What I learned: TODO: what was learned is not documented.

Demo — TODO: a demo URL is not published.

Personal / experimental

RAG System

Problem: A personal experiment in grounding model answers in a private document set.
Architecture: Documents, a retrieval step, context assembly, and a language model. The index and the model host are not specified.
Technology: Retrieval-augmented generation. TODO: vector store, embeddings, and model are not documented.
What I learned: TODO: what was learned is not documented.

Demo — TODO: a demo URL is not published.

Personal / experimental

AI Agent

Problem: A personal experiment with an agent that turns a natural-language request into a bounded action.
Architecture: A request, a tool boundary, and an action step. The tool catalog is not specified.
Technology: Agent loop. TODO: framework, model, and tools are not documented.
What I learned: TODO: what was learned is not documented.

Demo — TODO: a demo URL is not published.

Personal / experimental

Computer Vision

Problem: Recognize grocery or retail products from an image.
Architecture: An image in, a recognition step, and a product label out. The model family and the dataset are not specified.
Technology: Computer vision, Grocery and product recognition. TODO: model, dataset, and runtime are not documented.
What I learned: TODO: what was learned is not documented.

Demo — TODO: a demo URL is not published.

Personal / experimental

Voice / Multimodal Assistant

Problem: A personal experiment combining voice or another modality with an assistant.
Architecture: A modality in, a model, and a spoken or multimodal response. The pipeline is not specified.
Technology: Voice, Multimodal assistant. TODO: speech, vision, and model components are not documented.
What I learned: TODO: what was learned is not documented.

Demo — TODO: a demo URL is not published.