Sunil Pinnamaneni

About Sunil Pinnamaneni

I am passionate about differentiable and probabilistic programming. I have experience with foundation model applications, large language models (LLMs), large-scale vision and multi-modal models, automatic differentiation frameworks, neural network training, model evaluation, low-level CPU/GPU programming, benchmarking, parallel algorithms, SDEs/DAEs, computer algebra systems, time series analysis, deep learning compilers, and MLIR dialects. As a researcher and software engineer, I facilitated the transition of technologies from a research setting to a production environment: Python, CUDA C/C++, CPU C++, Ruby, Mathematica, JavaScript, TypeScript, and others /: Foundation model application frameworks, deep learning model development frameworks, REST API-based web and mobile platforms, MPI based HPC applications, symbolic computer algebra system programming, and deep learning compilers : SIMT, object-oriented, procedural, functional, reactive, and CAS pattern matching /: Deep Learning Design Patterns, Gang of four patterns, enterprise architecture patterns, and SOLID /: TensorFlow, Keras, PyTorch, JAX, Equinox, Flax, HuggingFace ecosystem libraries, LlangChain, LlamaIndex, NeMo, Triton, and others : Prompt engineering, PEFT, adapters, RAG apps, semantic embeddings, image embeddings, transfer learning, knowledge distillation, data augmentation, synthetic data generation, ontology augmentation, and description logic assisted LLM reasoning : CNNs, BERTs, T5, CoLT5, GPT, Sparse Transformer, Llama variants, Falcon variants, TimeGPT, LLaVA variants, various efficient transformers, and other architectures : protobuf, GraphDef, SavedModel, Grappler, XLA, XLA HLO, StableHLO, state_dict, Torchscript, PyTorch JIT Compiler, PyTorch-XLA, pickle.msgpack, JAXpr, LLVM, MLIR, PTX, and CPU/GPU/TPU architecture : LLVM, XLA, and tensor, linalg, mhlo, tosa, affine, scf, gpu, tpu, and vector dialects : Z3 : Mathematica, Maple, Reduce, Axiom, Macaulay2, Singular, and Cocoa

Experience

1. Machine Learning Researcher

   Stealth Startup

   Jul 2025 — Present · NY, US

   Our team works on quantum machine learning on concept classes that aren\'t PAC-learnable on classical computers.

Education

• The Wharton School

  Entrepreneurship Acceleration Program Certificate, Entrepreneurship/Entrepreneurial Studies

• CQF Institute

  Wilmott Certificate in Quantitative Finance, Quantitative Finance

• New York University

  PhD, Mathematics

• University of Chicago

  Bachelor of Science (BS), Mathematics