Mark K. Langat
I’m Mark K. Langat, a software engineer and researcher focused on neuro-symbolic AI, reinforcement learning, automated reasoning, and the security of cryptographic and quantum systems.
About
I’m Mark K. Langat, a software engineer and researcher focused on neuro-symbolic AI, reinforcement learning, automated reasoning, and the security of cryptographic and quantum systems. I build verification and test-time tooling in Rust and Python, with a strong emphasis on correctness, symbolic equivalence, and verifier-guided reward modeling.
Across industry and research, I’ve led and shipped systems that combine formal methods with machine learning—supporting randomized testing, conformance checking, and hallucination-mitigation approaches for language models. I’m especially interested in turning hard combinatorial decision problems into learning-guided workflows using MCTS, SAT/SMT reasoning, and conflict-driven clause learning.
- 11 years' experience
- Joined October 2026
Industry experience
- Software & Internet
- Computers & Electronics
- Financial Services
- Professional Services
- Government
Work experience· 5 roles
Developed syntactic data generation, test-time compute, and neuro-symbolic AI techniques using reinforcement learning with verifier feedback, semantic consistency, symbolic equivalence, and reward modeling to improve decision quality and correctness.
Led ML and AI work on program-security analysis of cryptographic code and quantum computers, focusing on learning randomized reductions and analyzing program properties in MATLAB, Pascal, and C/C++. Built Rust services and integrated them into the testing workflow of a distributed journal-management system.
Worked on mathematically sound automated reasoning checks for Amazon Bedrock Guardrails. Co-invented a U.S. patent filing on mitigating hallucinations in large language models, and applied neuro-symbolic AI to capture symbolic knowledge and reduce factual errors.
Education
Hire Mark K. Langat today
Post a job and invite Mark, or contact them directly. No hidden commission.