I'm a software engineer and researcher focused on AI reliability, distributed systems, and functional programming. I build infrastructure for LLM research on the Elixir/BEAM platform.

I'm the creator of the Crucible Framework, a platform for conducting reproducible experiments on large language model reliability, built on Elixir/OTP.

Key Goal: Building towards 99%+ LLM reliability through ensemble voting and request hedging, with comprehensive statistical testing and transparent causal reasoning chains.

All published under the @North-Shore-AI organization:

Tech Stack: Elixir, OTP, BEAM VM, Telemetry Research Areas: LLM reliability, ensemble methods, tail latency optimization, statistical testing Status: Active development, v0.1.0 released

• synapse ⭐ 23 - Synapse: Elixir-powered AI agent orchestration, built on the battle-teste…
• ds_ex ⭐ 14 - DSPEx - Declarative Self-improving Elixir | A BEAM-Native AI Program O…
• DSPex ⭐ 8 - Declarative Self Improving Elixir - DSPy Orchestration in Elixir
• mabeam ⭐ 4 - Multi Agent BEAM
• AutoElixir ⭐ 3 - AI Multi Agent Swarms in Elixir
• ALTAR ⭐ 4 - The Agent & Tool Arbitration Protocol

• gemini_ex ⭐ 18 - Elixir Interface / Adapter for Google Gemini LLM, for both AI Studio a…
• claude_agent_sdk ⭐ 7 - Elixir SDK for Claude AI Agent API - Renamed from claude_code_sdk_elix…
• codex_sdk ⭐ 1 - OpenAI Codex SDK written in Elixir
• jules_ex ⭐ 1 - Elixir client SDK for the Jules API - orchestrate AI coding sessions
• pipeline_ex ⭐ 6 - Claude Code + Gemini AI collaboration orchestration tools

• gepa_ex ⭐ 0 - GEPA (Genetic-Pareto) optimizer combining LLM-powered reflection with Pareto search to evolve text-based system components
• snakebridge ⭐ 0 - Configuration-driven Python library integration for Elixir - auto-generate type-safe wrappers for any Python library
• weaviate_ex ⭐ 0 - Modern Elixir client for Weaviate vector database with health checks…
• json_remedy ⭐ 20 - A practical, multi-layered JSON repair library for Elixir that intelli…
• snakepit ⭐ 8 - High-performance, generalized process pooler and session manager for e…
• duckdb_ex ⭐ 0 - DuckDB driver client in Elixir

• sinter ⭐ 9 - Unified schema definition, validation, and JSON generation for Elixir
• exdantic ⭐ 8 - A powerful, flexible schema definition and validation library for Elix…
• perimeter ⭐ 6 - Elixir Typing Mechanism

• ex_dbg ⭐ 9 - State-of-the-Art Introspection and Debugging System for Elixir/Phoenix…
• elixir_scope ⭐ 4 - Revolutionary AST-based debugging and code intelligence platform for E…
• ElixirScope ⭐ 3 - AI-Powered Execution Cinema Debugger for Elixir/BEAM
• elixir_dashboard ⭐ 0 - A Phoenix LiveView performance monitoring dashboard for tracking slow endpoints and database queries
• elixir_tracer ⭐ 0 - Local-first observability for Elixir with New Relic API parity

• superlearner ⭐ 6 - OTP Supervisor Educational Platform
• apex ⭐ 3 - Core Apex framework for OTP supervision and monitoring
• apex_ui ⭐ 3 - Web UI for Apex OTP supervision and monitoring tools
• arsenal ⭐ 3 - Metaprogramming framework for automatic REST API generation from OTP o…
• arsenal_plug ⭐ 2 - Phoenix/Plug adapter for Apex Arsenal framework

• supertester ⭐ 3 - A battle-hardened testing toolkit for building robust and resilient El…
• sandbox ⭐ 3 - Isolated OTP application management system for Elixir/Erlang
• cluster_test ⭐ 3 - Distributed Erlang/Elixir test cluster management via Mix tasks

• foundation ⭐ 10 - Elixir infrastructure and Observability Library
• AITrace ⭐ 0 - The unified observability layer for the AI Control Plane
• Assessor ⭐ 0 - The definitive CI/CD platform for AI Quality.
• Citadel ⭐ 0 - The command and control layer for the AI-powered enterprise

• cf_ex ⭐ 3 - Elixir libraries for Cloudflare edge computing services. Battle-tested…
• ex_cloudflare_phoenix ⭐ 0 - Cloudflare Durable Objects and Calls for Phoenix Framework

• playwriter ⭐ 6 - Elixir WSL-to-Windows browser integration

• youtube_audio_dl ⭐ 0 - Download high-quality audio from YouTube as MP3 files using Elixir. Fe…
• tools ⭐ 0 - Elixir repository

Languages: Elixir, Erlang, Python, JavaScript/TypeScript, Rust

Frameworks: Phoenix, OTP, FastAPI, React

Specialties:

• Distributed systems & fault tolerance
• AI/LLM infrastructure & reliability
• Functional programming & metaprogramming
• Statistical analysis & experimental design
• Developer tools & productivity

Platforms: BEAM VM, AWS, GCP, Cloudflare Workers, Edge Computing

• Research: LLM reliability through ensemble methods and statistical testing
• Building: AI infrastructure on Elixir/OTP
• Learning: Advanced OTP patterns, distributed systems optimization
• Growing: The Crucible framework ecosystem

• GitHub: @nshkrdotcom
• Organization: @North-Shore-AI

"Build infrastructure that researchers and engineers actually want to use. Make reliability measurable. Make experiments reproducible. Make the BEAM shine for AI workloads."

• Collaboration on Elixir AI tooling
• Consulting for distributed systems & AI infrastructure
• Speaking about LLM reliability, Elixir/OTP, or functional programming
• Research partnerships in AI reliability & distributed systems
• Open source contributions - PRs welcome on any project!

Period: September 28 - October 27, 2025

1. DSPex - 4.8M lines (18 commits) - Declarative AI orchestration framework
2. pump_fun_web - 200K lines (202 commits) - Solana memecoin platform
3. snakepit - 140K lines (127 commits) - Python-Elixir bridge & process pooling
4. claude_agent_sdk - 82K lines (65 commits) - Claude AI agent integration
5. Spectra - 70K lines (8 commits) - Web application project
6. code_agent - 38K lines (38 commits) - AI coding agent system
7. weaviate_ex - 29K lines (28 commits) - Vector database client
8. nshkrdotcom.github.io - 28K lines (219 commits) - Portfolio & documentation site
9. gemini_ex - 26K lines (23 commits) - Google Gemini LLM client
10. pipeline_ex - 24K lines (7 commits) - Multi-LLM orchestration tools

Period: September 28 - October 27, 2025

Models Used: Sonnet 4.5, Sonnet 4, Haiku 4.5, Haiku 3.5, Opus 4.1

High-Impact Work:

• 89% of changes focused on AI/LLM infrastructure and SDK development
• Major refactoring/enhancement of DSPex declarative AI framework
• Production web application (pump_fun_web) with 202 commits shows sustained delivery
• Parallel development across 44 repositories demonstrates broad ecosystem growth

AI-Assisted Development ROI:

• Total Investment: $3,347 over 30 days
• Output: 5.6M lines of code across 978 commits
• Efficiency: ~$1 per 1,000 lines of thoughtfully architected code
• Time Multiplier: 95% cache hit rate indicates highly efficient iterative development
• Context Leverage: Average 6.2M tokens/commit shows deep architectural work

Peak Activity Correlation:

• Oct 7-11: Infrastructure automation & AI framework development ($391-$256/day)
• Oct 14: Repository organization & documentation systems ($110/day)
• Oct 17-20: Multi-repo AI SDK development ($253-$182/day)
• Higher costs correlate with complex architectural decisions & framework design

Notable Achievements:

• Built/enhanced 10 AI orchestration & SDK projects
• Delivered production Solana platform with 200K+ LOC
• Created comprehensive Python-Elixir bridge (snakepit)
• Developed multiple vector DB & graph DB clients
• Maintained documentation hub with 219 commits
• Automated GitHub star tracking & README generation

Primary Activities:

• 89% AI/LLM infrastructure & agent orchestration
• 5% Production web applications
• 4% Data infrastructure & database clients
• 2% Developer tooling & automation

Development Velocity:

• Daily Average: 187,977 lines/day (32.5 commits/day)
• Per Session: 5,766 lines/commit
• With AI Assistance: 943 lines per dollar spent

Quality Indicators:

• High commit-to-LOC ratio (5,766 lines/commit) suggests substantial, well-planned changes
• 95% cache hit rate indicates consistent, iterative refinement
• Distributed across 44 repos shows ecosystem thinking vs. siloed development
• 6.2M tokens/commit demonstrates thorough architectural consideration

Last updated: 2025-11-06