SLURM-like topology-aware GPU scheduler for multi-GPU servers (single node).

custom, secure, isolated sandbox for DeepAgents, built without containers or VMs and adhering to the SandboxBackendProtocol.

Remote first postgres monitor, helps manage db on vps or remote pc easily, like GUI for postgresSQL.

code executor for running untrusted code, No containers, VMs, Native support for gpus.

understanding deep learning frameworks by implementing core components.

basic benchmarking kernel backends against a GPT model, from torch to Triton to raw CUDA.

RL environment for simulated GPU scheduling of mixed training and inference jobs.

openEnv Environment for RL-Based Code Optimization using perf, valgrind etc.

intelligent CI/CD pipelines with AI agents for automated testing and deployment.

building a GPT model from scratch to understand transformer architecture deeply.

building a compiler end to end, from lexing and parsing to IR design and code generation.

exploring CUDA and GPU computing for high-performance parallel processing.

understanding RAG by implementing Naive RAG, Multi Query, RAG-Fusion, Decomposition, Step-Back, HyDE, CRAG, and Self-RAG in langchain.

COBOL dataset generator for training and benchmarking.

Boilerplate code generator templates.