What changed in Praval between 0.7.22 and 0.8.1
How Praval gained a common model runtime, MCP tools, multimodal input, voice, stronger Agent coordination, and exact-wheel validation while preserving its original design.
Rajesh Sampathkumar - All Posts
AI, Aviation, Geopolitics, Philosophy, and more - A polymathic exploration
Latest posts
Building Vajra 3D Shape Search: Text Queries over Indexed Point Clouds
How Vajra 3D searches indexed point-cloud objects from text queries using a closed-source embedding model, Vajra HNSW, BM25 metadata search, RRF fusion, and a Three.js demo.
The Realities of AI, April 2026
There is a lot of noise in the present state of AI discourse. Some of it comes from legitimate technical excitement about the current state of AI systems. Be it coding or image/video generation or anything else. Some of it comes from venture capital, public markets, product marketing, and the prestige incentives of major AI labs. We have seen AI labs make stupid and exaggerated claims, and the…
The Agreeable Machine: AI Sycophancy and the Mind It Shapes
On what it does to the mind when the most sophisticated conversational systems we have ever built are trained, by design, to agree with you.
Human-AI Collaboration: Preserving Human Agency in the Age of Artificial Intelligence
Why we must hold onto friction, understanding, and prioritize personal growth when working with AI, instead of ceding our cognitive processes to machines.
Vajra Search with a Rust backend (v0.2.1): Greater Build and Latency Performance
Vajra Search is the successor to Vajra-bm25, with a new backend implemented in Rust that covers the vector index core and which is published to PyPI as v0.2.1.
From Vajra v0.4.1 to v0.5.0 to Rust backend: architecture notes and benchmark results
Technical notes on the Vajra search stack transition from BM25 benchmarks and Python vector search (v0.4.1/v0.5.0) to the Rust HNSW backend (`vajra_search`), with reproducible Wikipedia results.
Re-engineering Vajra's vector search with inspiration from ZVec
Vajra's HNSW vector search started 217× slower to build and 8× slower to query than ZVec. Six targeted engineering changes brought it to 36× and 1.6× respectively — without rewriting in C++ or adding Numba.
How to use AI: Radical Delegation versus Personal Thriving
A practical philosophy for avoiding the _dvandva_ using AI as leverage while still building judgment, skill, and character.
Vidai: Teaching Machines Arithmetic Without Teaching Arithmetic
A neural system that learns to parse mathematical notation rather than compute it, achieving 90%+ accuracy by separating structure understanding from calculation.
ToDACoMM: Topology and The Shape of Learning Algorithms
When you train a neural network, you're running a dynamical system that carves out a representation space. ToDACoMM measures the topology of what gets carved, revealing a striking divide between encoders and decoders, and now extends to MLPs and large-scale transformer analysis.
AI Explorations - 2025 In Review
A comprehensive look back at my year in AI - scaling up at work, building the Praval agentic AI framework, Vajra Search, Tlon mathematics, ToDACoMM, exploring mechanistic interpretability, honest research failures, and reflections on curiosity vs. accomplishment.
Deep-Lyapunov: Deep Learning Dynamical Systems Analysis
Applying perturbation analysis and Lyapunov exponents to neural network training. Dense networks converge; transformers diverge. The architecture determines the stability.
Vajra BM25: Building a Search Engine with Category Theory
After months of development, Vajra BM25 achieves ~1.2-1.3x faster latency than BM25S while maintaining competitive accuracy. I share what I learned building it and benchmark results across BEIR and Wikipedia datasets.
Tlön Mathematics: A Process-First Framework Inspired by Borges
Building a mathematical framework where processes are primitive and objects emerge as stable patterns - with 20 axioms, rigorous proofs, and simulations of dynamical systems that demonstrate the core insight: stability is special, not generic.
Praval Deep Research – Local-First AI Research Assistant
Building a privacy-focused research assistant with multi-agent architecture: architectural decisions, local-first design, and lessons learned from production deployment.
Praval Analytics – Reimagining Business Intelligence Through AI Agents
Why traditional BI breaks on factory floors, and how multi-agent architecture enables conversational analytics: architectural decisions, event-driven coordination, and production lessons.
Praval – Agentic AI Framework in Python
Over the last several months, I have been building and developing Praval, a Pythonic agentic AI framework for multi-agent system development inspired by coral ecosystems.
Problem Frontier Expansion: What the Age of AI Deserves
Peripatetic wandering is not a phrase I would normally associate with work, and yet, as of the end of 2025, I do. The truth is that the work of the past isn't the work of today.
Artificial Intelligence and Human Creativity
Many interesting narratives populate the philosophical discourse around artificial intelligence, not the least of which is the potential of artificial intelligence to replace humans in different endeavors.