Welcome! This GitHub profile is the home of my open-source experiments and research projects around agentic AI, machine learning, search and information retrieval, and emergent dynamics in large language models.

• Principal Research Focus: Multi-agent systems, mathematical foundations of AI, graphs, dynamical systems, and interpretable machine learning.
• I love building small, testable research codebases that bridge mathematical foundations to hands-on implementations—especially if it's agentic, compositional, or reveals deep or interesting dynamics.
• My background mixes AI, math, distributed systems, and software engineering from recent work experiences; and I always favor clarity, reproducibility, and simple well-documented demos.
• Past work experience includes mechanical engineering and aerospace design, manufacturing, quality management and statistical problem solving
• I've worked across industry domains such as automotive and truck manufacturing, aerospace, telecommunications, renewable energy, and product SaaS (data platform products, HR tech products)
• Portfolio Website: rajeshrs.in
• LinkedIn: linkedin.com/in/rajeshrs

• Praval — Multi-Agent AI Framework
  Pythonic framework for building collaborative multi-agent systems. Compose simple agents for emergent capability, with built-in conversation memory, inter-process protocol (“Spore”), and LLM adapters.

  • Highlights: agent decorator, tool use, composable API, multiple transport layers (RabbitMQ), memory integrations, code extensibility.
  • Repo

• Praval Deep Research — Local-first AI Research Assistant
  Local-first research assistant for arXiv papers (runtime search, indexing, retrieval, agentic workflows and math parsing), built atop Praval.

  • Repo

• Vajra BM25
  Fast, vectorized BM25 search engine with category theory abstractions. Includes optimized index formats and parallel query engine.

  • PyPI: vajra-bm25 (current version 0.2.1)
  • Repo

• Tlon Mathematics
  A Python library for mathematical process theory—where “processes” are primitive and “objects” emerge as stable patterns. Features dynamical systems demos (double pendulum, N-body, Keplerian motion).

  • Repo

• Agentic NLD — Chaos Theory in LLM Conversations
  Experiments with two-agent LLM conversations exhibiting chaotic signatures and sensitive dependence on initial conditions (discrete-time dynamical system for language).

  • Repo

• Most projects are Python-first (with some TypeScript, Shell, and TeX/PDF/Markdown documentation)
• I organize repos by research question and try to keep everything well-tested, simple, and reproducible.
• Contributions: Open to issues, PRs (prefer discussions first).

General setup for any project:

python3 -m venv venv
source venv/bin/activate
pip install -e ".[dev]"   # when available
pytest -v

Check each repo’s README for project-specific install/demos.

• Open to collaborations on agent-based AI, interpretable ML, mathematical frameworks, and high-performance search.
• For PRs, open an issue or discussion first; I value alignment and thoughtful reviews.
• MIT license by default; cite my work if you use it for research (I enjoy seeing derivatives).

• GitHub Profile
• Website: rajeshrs.in
• LinkedIn: linkedin.com/in/rajeshrs
• Twitter
• For questions/collaboration: open an issue and tag me

• 📫 Email
• 🐦 Twitter
• 💼 LinkedIn