Building Intelligent Systems

Mapping the latent spaces of data. My major interest lies in deciphering neural architectures to build the next generation of autonomous ecosystems.

previous work

Origins Story

Rakesh Varma Sirivuri

I am a Btech student driven by passion for Neural Networks and computer vision. My work focuses on exploring the underlying "why" of complex architectures, pushing the boundaries of current AI research to invent novel ideas and transformative solutions. By bridging the gap between theoretical exploration and practical implementation, I strive to develop robust, real-world applications that solve intricate problems and contribute to the evolution of intelligent systems.

Experience

Reseearch Intern at HCI lab IIITS

During the summer of 2026, Introduced a new mathematical framework for optimising the rendering of the 3DGS(reconstruction of 3D models using 2D images).

PyTorch CUDA Gaussian Splating

Featured Projects

Landslide MoE Predictor

A deep learning model utilizing a Domain-Decoupled Mixture of Experts (MoE) architecture to predict landslide susceptibilities across varying geographical domains.

PyTorch Python Streamlit

Autonomous navigation rover

A self-navigating rover featuring real-time obstacle avoidance using computer vision and ultra sonic for complex terrains.

Opencv Raspberrypi Yolo_v8

CodeSense-Doc-string generator

An Intelligent agent that automatically generates professional Google-style docstrings and grades the maintainability of your Python code in real-time.

LLMs FASTAPI LANGCHAIN

Open-Source Chatbot

A versatile, locally-deployable conversational AI built with open-source models. Features context retention, RAG capabilities, and a responsive UI.

Gemma 3 Python Hugging Face

Cardiac Risk Predictor

A machine learning pipeline trained on clinical datasets to evaluate patient metrics and predict the probability of heart disease, aiding in early diagnosis.

Sci-kit learn Decision Tree's AWS

Smart Study Buddy

A local NotebookLM which summarizes PDF using qwen3

QWEN Hugginface Langchain

The Bumblebee Principle

"According to all known laws of aviation, the bumblebee should not be able to fly. Its wingspan is not in right proportion to its body mass, but ignoring all these laws it flies anyway."

For decades, traditional computing paradigms suggested that modeling human-like intuition or parsing hyper-complex, non-linear data was impossible. But just like the bumblebee, Neural Networks don't care about the constraints of classical logic. They find the underlying patterns anyway.

My engineering philosophy is rooted in this persistence. I design deep learning architectures and autonomous systems to challenge the boundaries of what machines can achieve—pushing past the "impossible" until it takes flight.

Transmit Signals

I am actively seeking new challenges and collaborative ventures in the tech sector. Whether you have a project in mind, a specific question, or simply want to connect, feel free to send a signal across the void!

Send Signal