Askar Uvaliyev announced a $3 million pre-seed funding round for Riasm. The investment syndicate includes venture funds Paradigm and SV Angel, prominent artificial intelligence researcher Fei-Fei Li, and Eurasian Hub founder Nurdaulet Bazylbekov. The startup is engineering algorithmic models designed to detect flaws and unsubstantiated assumptions within scientific reasoning.
The founder disclosed the transaction details to Digital Business on October 5. According to Uvaliyev, the company was incorporated in early September 2026, and he currently operates as the sole engineer on the venture.
Validating Underlying Engineering Assumptions
Riasm’s targeted customer profile encompasses deep-tech research and engineering startups in robotics, computational physical simulation, and AI compute infrastructure. Uvaliyev articulates two primary problem sets: diagnosing why a complex physics simulation diverges from physical reality, and stress-testing whether robotic locomotion control policies can achieve greater operational resilience.
According to the founder’s technical brief, Riasm is engineered to audit and isolate vulnerable assumptions underpinning mechanical and algorithmic engineering models.
On its official landing page, the company categorizes its capabilities across domain-specific model fine-tuning, data engineering, and computational physical sciences. The prototype software is designated as the Riasm Supercomputer, available via referral-gated private pilots. This represents self-reported developer positioning; independent benchmark evaluations are not yet publicly available.
Technical Background in Mathematical AI Foundations
Prior to Riasm, Uvaliyev contributed to the creation of the Nomos-1 mathematical reasoning model developed by Nous Research, according to Digital Business. The open model repository confirms Askar / massiveaxe among the accredited co-authors. Nomos-1 is engineered to solve formal mathematical problems and generate natural-language mathematical proofs.
This engineering background centers on formal domain verification where mathematical correctness can be proven analytically. However, Nomos-1 and Riasm represent distinct software architectures: performance on mathematical benchmarks does not automatically validate systemic reliability when evaluating multi-variable physical scientific literature.
Post-Round Execution Milestones
Riasm is currently conducting closed pilots with early-stage technical startups in San Francisco. Client identities and revenue metrics remain undisclosed. The newly secured capital will be allocated toward technical recruitment, GPU cluster compute access, and expanding customer pilots, with plans to open-source select model checkpoints down the line.
RAEM.KZ previously profiled Arlan Rakhmetzhanov and Nozomio, an AI startup building specialized contextual memory infrastructure for autonomous agents. Riasm contributes an adjacent deep-tech perspective: algorithmic verification of fundamental axioms in scientific research.
Assessing the long-term utility of the product will require verifiable pilot results: what structural flaws the model detects, its false-positive error margins, and the quantifiable research velocity gained by enterprise engineering teams. A venture funding announcement alone does not answer those operational questions.
