3. Data, Science & AI

Optimizes Python numerical computations via C extensions for scientific computing and portfolio risk/return calculations, targeting matrix operations and performance bottlenecks.

Guides implementation of tensor parallelism in PyTorch for column-parallel and row-parallel linear layers in distributed neural networks.

Manages multiple Letta AI agents declaratively using a kubectl-style CLI for configuration, memory, tools, and folder management.

Guides writing SPARQL queries for RDF/Turtle datasets, supporting filtering, aggregation, and Turtle file operations in academic contexts.

Guides minimal GPT-2 inference implementation in constrained environments, covering checkpoint parsing, BPE tokenization, and common pitfalls for code golf challenges.

Guides optimization of eigenvalue computations for small dense matrices, achieving faster performance than NumPy/SciPy via Cython and LAPACK.

Guides search for probability distributions meeting exact statistical constraints, including KL divergence targets and high-dimensional optimization.

Provides guidance for optimizing slow SQL queries through plan analysis, benchmarking, and database-specific techniques to enhance performance.

Guides deployment of HuggingFace models as REST APIs using Flask or FastAPI for NLP inference tasks like sentiment analysis and text classification.

Guides recovery of data from corrupted or truncated SQLite databases using binary analysis and manual parsing techniques for damaged files.

Guides fusion protein design for FRET biosensors, including spectral analysis, codon optimization, and PDB database integration for gene synthesis.

Guides semantic similarity retrieval using embedding models (MTEB benchmarks) for document ranking and top-k item identification.

Extracts and processes data from ELF binary files, converting memory contents and headers into structured JSON for analysis.

Processes financial documents via OCR, extracts data, and generates structured outputs (CSV/JSON) while ensuring data safety.

Extracts text commands and typed input from video recordings using OCR, suitable for text-based games, terminal sessions, and screen recordings.

Guides primer design and DNA assembly simulation for Golden Gate cloning and Type IIS restriction enzyme workflows.

Guides construction of discrete/continuous probability distributions meeting statistical constraints via numerical optimization, including KL divergence and entropy targets.

Guides implementation of tensor parallelism in PyTorch, including sharding linear layers across GPUs and simulating collective operations like all-gather and all-reduce.

Guides Raman spectrum peak fitting with Lorentzian/Gaussian models, emphasizing data verification and physical constraints for material analysis.

Guides peak fitting in Raman spectroscopy data for materials like graphene, using Lorentzian, Gaussian, or Voigt functions for G, 2D, and D peaks.

Creates standalone CLI tools for neural network inference by reimplementing PyTorch models in C/C++, eliminating Python dependencies.

Guides DNA assembly techniques including Golden Gate, Gibson, and restriction enzyme cloning, covering primer design for Type IIS enzymes and fragment assembly strategies.

Guides reverse-engineering of ray-traced images by determining rendering parameters and implementing path tracing algorithms to match target images with high similarity.

Guides extraction of weight matrices from ReLU neural networks via input-output queries for model reverse-engineering and parameter recovery.