ThinkingDBx builds the Bonacci family of agentic data engineering & data science platforms, powered by Bonacci MoE 9B, our purpose-built mixture-of-experts model for DE/DS workflows. We also build ThinkingLanguage, the first compiled language for data & AI, and ThinkingMemory, agent-agnostic memory infrastructure.
ThinkingDBx is a technology startup based in Hyderabad, India, building the next generation of intelligent data infrastructure. At the heart of our ecosystem sits Bonacci MoE 9B, the first purpose-built mixture-of-experts model for Data Engineering & Data Science. With 9B total parameters and only 2.6B active per token across 8 experts and top-2 routing, it delivers expert-level SQL generation across 5 dialects, PySpark pipeline authoring, data modeling, and autonomous multi-step orchestration at small-model cost.
The model powers the Bonacci family of platforms. Bonacci Studio is an agentic data engineering platform where pipelines are built visually or through AI chat. Bonacci Flow is a pure agentic data engineering & data science platform that turns high-level goals into deployed, self-healing pipelines through Pipeline Autopilot.
We also build ThinkingLanguage, a compiled programming language where data pipelines, ML, and streaming are first-class features, and ThinkingMemory, an open-source, agent-agnostic memory platform that gives AI agents persistent, layered memory across sessions, tools, and teams.
We build for builders. From solo developers to enterprise teams, our tools are designed to eliminate complexity and make working with data 100x faster.
An AI-driven data engineering and streaming platform that helps businesses build, manage, and scale data infrastructure 100x faster — with Agentic, visual, no-code pipelines.
Visual pipeline designer powered by Apache Spark, Kafka & Camel. Agentic data engineering through AI chat — build pipelines with Spark, SQL, PySpark, MCPs, Kafka & JDBC.
A pure agentic data engineering & data science platform — powered by Bonacci, our fine-tuned AI model purpose-built for DE/DS workflows.
No drag-and-drop, no canvas — just intent. Agents plan, build, train, and ship pipelines and models end-to-end, with the Bonacci model reasoning natively over Spark, SQL, PySpark, and ML frameworks.
The first purpose-built mixture-of-experts model for Data Engineering & Data Science. Expert-level SQL generation, PySpark pipeline authoring, data modeling, and autonomous multi-step orchestration — all in a 9B sparse model that activates only 2.6B parameters per token.
Trained on 122B+ tokens (general corpus + FineWeb-Edu) on TPU v6e-16, post-trained with SFT + GRPO on 15K curated DE/DS pairs. Native first-class support for ThinkingLanguage pipelines.
The world's first compiled language where data pipelines, SQL-like queries, ML training, and real-time streaming are all first-class language features — not libraries bolted on after the fact.
Compiled to native code via LLVM & Cranelift. Built entirely in Rust. Python-like readability meets Rust-like safety and performance.
Agent-agnostic memory platform with a layered architecture — Working, Episodic, Semantic & Procedural memory. Reasoning-aware retrieval with built-in forgetting & compression.
Make AI agents 95% cheaper, consistently reliable, and genuinely improving over time. Works with any framework or LLM.
Deep dives into data engineering, AI agents, and platform architecture. From schema drift detection to autonomous monitoring — explore the ideas behind our products.
Technical articles, product announcements, and industry analysis for builders and engineers.