Repositories
Deva-1903/medic-self-healing-deploys
A self-healing deploy watchdog: detects a broken deploy, repairs a bounded set of faults, and proves every repair in an isolated copy first. The LLM may propose; only the deterministic gate may promote.
Deva-1903/ccr-platform
Deva-1903/resume-pipeline-starter
A Claude Code pipeline that turns a job description into a truthful, tailored, one-page LaTeX resume. Truth-audited, ATS-aware, and it tracks the skills you are missing.
Deva-1903/Deva-1903
Config files for my GitHub profile.
Deva-1903/provisioning-platform
Distributed server provisioning & configuration control plane — Go, gRPC, PostgreSQL, Redis
Deva-1903/gpu-scheduler
Multi-tenant GPU job scheduler in Go: weighted DRF fair sharing, best-fit bin-packing, priority preemption. Kubernetes + simulated executors, Postgres + Redis, benchmarked with a Python load harness.
Deva-1903/neon-quantized-gemm
Hand-written INT8/INT4 GEMM kernels for ARM NEON sdot on Apple M4 - fp32 baseline to 4x4 register-tiled micro-kernel (132x), bit-exact & benchmarked p50/p99/p999
Deva-1903/meshery
Meshery, the cloud native manager
Deva-1903/cs690pf
CS 690PF Performance Engineering — assignments 0-2 and final project
Deva-1903/ciir_agentic_search
Deva-1903/Spark-ETL-Optimization
Deva-1903/refusal-decay
Deva-1903/ClaudeResumePipeline
Deva-1903/cpp-autodiff-engine
Deva-1903/WeekFlow
Deva-1903/spotiplay
Deva-1903/ngvi-curvature-variance
Deva-1903/cf_ai_research_scout
Deva-1903/Expense-Tracker
Deva-1903/AutoEval
Deva-1903/autoresearch-karpathy
AI agents running research on single-GPU nanochat training automatically