Arnav Mehta

Co-Founder

Berkeley, California, United States2 yrs 7 mos experience

Key Highlights

Expert in machine learning and formal methods.
Founder of an innovative AI tutoring company.
Researcher developing advanced educational tools.

Stackforce AI infers this person is a Machine Learning Researcher with a focus on Education Technology.

Contact

Skills

Core Skills

Machine LearningFormal MethodsEducationNumerical AnalysisGame Development

Other Skills

Deep LearningLean 4Automated Proof EvaluationDataset CreationBenchmarkingProject-Based LearningLLM-powered ToolsVideo Lesson GenerationStochastic Time SeriesNeural NetworksDimensional ReductionDiffusion MapsAutoencoder AlgorithmsPythonUnity

About

I am a machine learning researcher at UC Berkeley, working at the intersection of AI, formal methods, and education. My research focuses on deep learning for automated proof evaluation. Further, I lead Sonder, a company building AI tutors that teach visually.

Experience

2 yrs 7 mos

Total Experience

1 yr 3 mos

Average Tenure

1 yr 4 mos

Current Experience

Sonder

Founder

Jan 2025 – Present · 1 yr 4 mos · San Francisco Bay Area · On-site

Taught 100+ high school students Machine Learning through one-on-one, project-based learning. Now building LLM-powered tools to enabling teachers to generate visual video lessons directly from text.

Machine LearningProject-Based LearningLLM-powered ToolsVideo Lesson GenerationEducation

University of california, berkeley

Researcher

Aug 2024 – Present · 1 yr 9 mos · Berkeley, California, United States

Building LeanTutor, an LLM-powered math tutor that automatically formalises student-written proofs into Lean 4, verifies their correctness, and provides pedagogical feedback. Co-created PeanoBench, a benchmark dataset for evaluating auto-formalisers in the educational settings. Our work is supported by a grant from the AI for Math Fund by Renaissance Philanthropy and XTX Markets.

Machine LearningDeep LearningFormal MethodsLean 4Automated Proof Evaluation

University of maryland

Researcher

Jun 2023 – Apr 2024 · 10 mos · College Park, Maryland, United States · On-site

NSF Fellow, REU under Prof. Maria Cameron (UMaryland):
Numerical analysis of stochastic time series to study events that are rare on the timescale of the system. Researched the applications of neural networks in dimensional reduction, to study the dynamics of chaotic stochastic systems. Developed modifications to diffusion maps and autoencoder algorithms to achieve optimal low dimensional embeddings of the system.

Numerical AnalysisStochastic Time SeriesNeural NetworksDimensional ReductionDiffusion MapsAutoencoder Algorithms+1