Skip to content
View lonexreb's full-sized avatar
💭
Midnight shipper!
💭
Midnight shipper!

Block or report lonexreb

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
lonexreb/README.md

Hey, I'm Shubhankar 👋

Building Computer Vision infrastructure & AI-native Sports Analytical tools at Paloa Labs in Palo Alto, CA.

I'm obsessed with making AI agents that actually work — event-driven systems, reinforcement learning from human feedback, and tools that turn messy real-world interfaces into clean APIs.


🔨 What I'm Building

Developer Tools & Platforms

Project What it does
🌐 site2cli GitHub stars GitHub forks Turn any website into a CLI/API for AI agents — v0.6.0 on PyPI with crawl, monitor & screenshot commands, LLM-powered structured extraction, proxy support, OAuth device flow, multi-site orchestration, browser sessions, workflow recording, background daemon, unified MCP server, 500+ tests passing
🔬 MCPstudio The Postman for MCP — visual platform to create, test & manage MCP servers. FastAPI + React + Zustand + React Query + shadcn/ui, DDD backend, feature-sliced frontend, connection wizard, tool execution engine, settings, tools library, resources/prompts browsers, onboarding tour, execution history & server discovery
🧩 zigsaw AI workflow automation platform — chat or drag-and-drop to build workflows. Multi-AI (Claude, GPT, Gemini, Groq), instant API deployment, 100+ integrations
🔭 kalam Terminal agent for scientific research, chip design, hardware engineering & systems security — built on Deep Agents SDK. MCP tool servers: arXiv, Yosys, ngspice, nmap, Ghidra, Icarus, GTKWave, SymPy. Named after Dr. APJ Abdul Kalam
🔄 mobius High-performance framework for autonomous ML experimentation — Rust + Go + Python
🌟 morningstar Autonomous coding agent that turns Notion PRDs into working code — reads requirements, analyzes codebases, generates implementations with tests, commits & sends Slack progress updates. Powered by Claude Code
🧱 deepagents Batteries-included agent harness — LangChain + LangGraph with planning, filesystem ops, shell access, sub-agent spawning & auto-summarization. pip install deepagents
🐟 aish Claude Code plugin for the official Anthropic plugin marketplace — GPU cloud control plane (TensorDock + Modal MCP), GPU detection skills & ML workflow agents
🔭 compgit GitHub commits, quietly visible — across Chrome, iPhone, and macOS

AI Agent Systems & RL

Project What it does
🐙 tentalis ADHR meta-RL framework — AI agents that continuously learn from manager feedback via GRPO/CISPO/DAPO training, NATS messaging, process reward models & HaluGate hallucination scorer
RLYX-enhancer Hackable GRPO training framework with high-speed weight sync for multi-node RL
🤖 AGI-INC HALO-Agent — browser automation agent for AGI Inc's REAL Benchmark (112 web tasks across 11 sites). v5 with Gemini 3.1 Pro, scoring 38.5% on zero-site tasks
🎓 university-sim Multi-agent university simulation benchmark with Concordia GM + OpenClaw coordination + emergence metrics
🌸 flwr-frl-kit Federated RL Kit for flower.ai — contributing federated reinforcement learning to the Flower ecosystem
📈 DeepFinRL-UMass Multi-agent RL for automated stock trading & personalized ETF curation — hierarchical MARL with Director Agent + specialist analysts, LLM sentiment, PPO/CPPO/GRPO algorithms
🔬 autoresearch AI agents running research on single-GPU nanochat training automatically — bio/physics-inspired search loops (Lévy flights, Metropolis-Hastings, hippocampal replay, golden-angle sampling) for Karpathy's nanochat
🧮 claude-code-math-skills Claude Code harness for rigorous math problems — 7-agent Generator–Critic loop, Lean 4 formalizer, SymPy verifier, 334-chunk Lebl analysis corpus, Putnam/IMO benchmark suite

F1 & Motorsport AI

Project What it does
🏎️ aero-agent CFD-aware research assistant for F1 aerodynamicists — synthesize literature, correlate CFD, optimize every simulation hour under FIA ATR constraints
🔍 sim-scout Automated literature-to-simulation pipeline for automotive CFD — from published paper to benchmarked OpenFOAM simulation in hours, not weeks
🏁 f1-design-ai AI-assisted F1 car aerodynamic design pipeline — Blender parametric geometry + OpenFOAM CFD + ParaView analysis, orchestrated by OpenClaw

Founder Performance & Health

Project What it does
🧬 FounderAgent Autonomous health-aware AI agent — reads your biometrics, blocks calendar, sets Slack DND, sends recovery protocols
🏆 FounderLeague Gamified league platform where founders compete on readiness, sleep quality & recovery — anti-hustle-culture leaderboards
🧠 GameDay-AMS Cognitive Performance System for Founders — sports science meets knowledge work
🔋 FocusFuel-API Developer API that transforms wearable health data (HRV, sleep, training load) into cognitive performance predictions — 35 tests passing
🩺 AgentKit-Health Marketplace of pre-built AI agent templates for health workflows — Sleep Coach, Burnout Detector, Recovery Agent & more
💪 grithub Sports & healthcare platform with physiological simulation — FastAPI + BioGears integration for training plan generation. Hacklytics 2025 winner

Applied AI & Data

Project What it does
🏀 paloa-claw Personal basketball AI training assistant
🏥 clinical-trials-api ETL pipeline & REST API for ClinicalTrials.gov data — PostgreSQL, live on Render, built for OpenAlex integration
📧 sdr-automation Multi-step agent pipeline for SDR research and outreach — YC AI Hackathon project, built on Zigsaw
📺 zigsaw-labs AI-powered video ad creation & optimization — auto-generates, tests, and optimizes short-form video ads across TikTok, IG Reels & YouTube Shorts
📝 grammer-paraphrase-system Grammar correction & paraphrasing system with clean DDD architecture
🏠 HandyHommieAI AI-powered home appliance command center — upload manuals, ask questions, auto-call manufacturer support
🎯 growthclaw Autonomous founder-scouting outbound engine for Crowdstake AI — 3-stage pipeline (discover, enrich, draft), scrapes Product Hunt/Reddit/IH, scores leads 1-10, ~$0.02/lead. Built with OpenClaw at Austin Hackathon
🔲 kalam-placer Chip design macro placement optimizer — Partcl + Hudson River Trading competition ($20K prizes). Balances routing congestion, timing, power delivery & area on ibm01 benchmark

Fun & EdTech

Project What it does
♟️ dolphine-chess Multi-tenant SaaS chess academy management platform
🍁 Maple Emotion-aware AI tutoring companion — Hume EVI for emotion detection, DataBricks (DBRX + LanceDB) for RAG, Midnight blockchain for student data integrity. Smart topic breakdown & empathetic practice quizzes
🃏 blackjack-for-toddler Toddler-friendly Blackjack that teaches numbers & addition — React PWA with Capacitor for iOS
🃏 poker-for-toddler Texas Hold'em for kids (ages 4-9) with toy cars, gems & crowns — React PWA on Vercel
🕉️ Mahamrutyunjay-Mantra-Counter macOS desktop app for tracking Mahamrutyunjay Mantra chanting — voice detection, visual progress 1-108
🇮🇳 xplncert Inventors themselves walk you through NCERT — AI persona videos for Indian class 6–12 textbooks

🌍 Open Source Contributions

Shipping fixes, language support, and config improvements upstream — selected PRs:

openclaw/openclaw — agentic coding platform

📝 Workflow notes & lessons learned from this contribution sprint →

PR Status What it does
#74638 ✅ merged Accept browser.tabCleanup keys in zod schema
#75476 open Honor model.compat.unsupportedToolSchemaKeywords for OpenAI-completions tool schemas
#75445 open Tolerate unresolved SecretRef tokens during Discord/Telegram channel-actions discovery
#75425 reviewed Fall back to qrcode/lib subpath when bare specifier fails to resolve
#75339 open Normalize structured delta.content blocks to prevent [object Object] in chat replies
#75248 open Reorder workspace AGENTS.md template so load-bearing rules come first
#75217 open Honor skipBootstrap at the runtime injection path
#75210 reviewed Honor OPENCLAW_STATE_DIR for exec-approvals.json
#74945 open Canonicalize --model to lowercase before dispatch
#74921 reviewed Skip web tool discovery when surface is explicitly disabled
#74891 open Skip API-key prompt when user skipped installing the skill
#74643 open Per-agent verboseDefault and elevatedDefault config overrides
#74627 reviewed Clarify modelFallback semantics in deprecation warning

warpdotdev/warp — agentic terminal

📝 Workflow notes & lessons learned from this contribution sprint →

PR Status What it does
#9558 ✅ merged Align OSS .desktop Exec with packaged binary name
#9563 ✅ merged Attribute Alacritty/vte derivative code in two more files
#9670 ✅ merged Route CLIAgent::Pi to the default session listener
#9712 open Register Rename Active Pane as a keyboard-bindable action
#9568 open Add JSON language support via vscode-json-languageserver
#9562 open Add PHP language support via Intelephense LSP
#9560 open Strip linked-worktree + marker from branch picker
#9667 open Recognize Mistral Vibe as a CLI agent
#9669 open Fail fast on bootstrap when Node.js / yarn are missing
#9671 open Clear permission-scoped state when leaving the permission flow

NVlabs/alpamayo — NVIDIA's open foundational driving model (Physical AI / AV)

Hardening the public-facing surface — input validation, error UX, attention-impl correctness, and onboarding docs.

PR Status What it does
#73 ✅ merged Replace assert with ValueError in load_physical_aiavdataset() so input validation survives python -O
#74 open Show actionable error when HuggingFace dataset access is missing — surface gated-repo / 401 / 403 cases instead of confusing IndexError (fixes #59, #61)
#75 open Pin expert decoder to SDPA so the model loads when Flash-Attention 2 is globally enabled (fixes #52)
#76 open Clarify coordinate-frame conventions for project_waypoints_ftheta (refs #34)
#77 open Replace assert with ValueError across public-API input validation
#78 open Remove deprecated local_dir_use_symlinks arg from snapshot_download() call
#79 open Expand Troubleshooting with HF auth, FA2 cu_seqlens, and smoke tests for first-time users

Also contributing to

  • karpathy/autoresearch — bio/physics-inspired search strategies (Lévy flights, Metropolis-Hastings, golden-angle sampling, hippocampal replay) for the nanochat training loop
  • ourresearch/openalex-overview — OpenAlex architecture diagram generation script

🧠 Areas of Focus

  • AI Agent Systems — event-driven architectures, multi-agent coordination, tool use, autonomous coding agents
  • Reinforcement Learning — GRPO, DAPO, CISPO, process reward models, RLHF, hallucination scoring
  • Developer Tools — MCP servers, CLI generators, API discovery, workflow automation, agent harnesses
  • Applied ML — federated RL (flower.ai contributor), deep finance RL, chip design optimization
  • Founder Performance — biometric-driven agents, cognitive performance systems, gamified health optimization
  • Health & Wellness AI — agent templates for wearable data, sleep coaching, burnout detection
  • F1 & Motorsport AI — CFD-aware research assistants, aerodynamic optimization under FIA constraints, literature-to-simulation pipelines
  • EdTech — emotion-aware tutoring, gamified learning apps for toddlers, math through card games

🛠️ Tech Stack

Python TypeScript Rust Go Swift React FastAPI PyTorch LangChain NATS Pydantic Docker Playwright Next.js Supabase MongoDB PostgreSQL OpenFOAM

📫 Find Me

Portfolio Zigsaw LinkedIn


"Ship beats perfect" — I build things that solve real problems, then open-source them.

Pinned Loading

  1. site2cli site2cli Public

    Turn any website into a CLI/API for AI agents

    Python 20 6

  2. tentalis tentalis Public

    Managers managing AI agents with appraisal-style feedback loops using RLHF/GRPO

    Python 2

  3. flwr-frl-kit flwr-frl-kit Public

    Developing the RL KIT for flower.ai

    Python 1

  4. university-sim university-sim Public

    Multi-agent university simulation benchmark: Concordia GM + OpenClaw coordination + Moltbook emergence metrics. Compare LLM providers on agent coordination.

    Python 1

  5. AgentKit-Health AgentKit-Health Public

    Marketplace of pre-built AI agent templates for health workflows. Deploy Sleep Coach, Burnout Detector, Recovery Agent & more in minutes.

    TypeScript

  6. HandyHommieAI HandyHommieAI Public

    AI-powered home appliance management platform — upload manuals, ask questions, track warranties, auto-call support

    Python