ANLYTIC COMMUNITY • HACKING ON LOCAL LLMS, RAG & AI AGENTS
We don't just talk about AI.
We build it, break it, and ship it together in the room.
MEETUPSBi-Weekly Hack Nights
CODE100% MIT / Open Source
COMMUNITY2,500+ Friendly Builders
$anlytic run --lab 04-speculative-decoding --precision nf4
> Allocating 1x RTX 4090 (24GB VRAM) • Base model: Qwen-2.5-Coder-7B
> Initializing 1.5B speculative draft verifier with chunked prefill
> Speculative decoding verified: 168.4 tok/s throughput (2.4x speedup vs single model)
> Checkpoint exported to GGUF • Ready for Hack Night #04 live room demo [0 errors]
2,500+Community BuildersEngineers, researchers, and weekend hackers across 35+ countries
3Hackathons CompletedPacked in-person labs with 100% live code and running pipelines
48+ hrsLive Hacking & CodeHands-on build time, lightning demos, and peer architecture reviews
0Vendor Pitch DecksZero sales presentations or product sponsorships allowed on stage
Grab your laptop, clone the verified starter repo, and build something functional before the night ends. Friendly peers, local weights, and zero gatekeeping.
Read our complete origin story & ethosCOMMUNITY SHOWCASE • RECENT BUILDS
projects shipped by our members:
Hack Night #03
Agent Tool Calling96/96 Shipped Valid JSON
Hack Night #02
Hybrid RAG PipelineRecall 94.2% • 180ms p95
Hack Night #01
Local 8B Fine-Tuning65 Checkpoints Deployed
Next Hack Night
Hack Night #04: On-Device AI
Speculative decoding & vision-language models on your own hardware.
120 Seats • August 28, 2026Upcoming MeetupRSVP Open on Luma
HOW WE LEVEL UP
three ways we build together:
hack nights & labs
Live coding together with laptops & pizza
- Fine-Tuning Models on Consumer GPUs (Unsloth & LoRA)
- Deploying Fast vLLM & Ollama Local Engines
- Building Multi-Agent Workflows with LangGraph
- Hybrid Vector Search & Cross-Encoder Reranking
- Local Real-Time Voice & Vision Pipelines
- Community Lightning Demos & Peer Code Reviews
- Synthetic Dataset Bootstrapping & Filtering
- Hardware-Aware VRAM Allocation & Quantization
war stories
Unfiltered production failure breakdowns
- How Our Agent Looped and Ran Up a $14K Cloud Bill
- Debugging 14-Second P99 Latency Spikes in Production
- Silent Retrieval Failures in Production Vector Databases
- Why 512-Token Chunking Severed Our Financial Tables
- Handling LLM Non-Deterministic Hallucinations at Scale
- Building Real-Time Production Guardrails with Pydantic
- PyTorch GPU Memory Leaks & CUDA Cache Fragmentation
- Embedding Model Schema Drift That Broke Search Recall
paper sprints
Dissecting fresh preprints into code
- Speculative Decoding Implementations with Draft Models
- Mamba & Selective State Space Linear Architectures
- Vision-Language Model Native Function Tooling
- Test-Time Compute & Monte Carlo Search Reasoning
- Mixture of Experts (MoE) Top-K Routing & Gating
- Direct Preference Optimization (DPO) Tuning Loops
- Cross-Modal Weight Projection & Audio Spectrograms
- Quantization Noise & FP8 vs INT4 Activation Limits
ANLYTIC ECOSYSTEMtools we build &
tools we build &
hack with:
Open-weight models, high-performance serving runtimes, and local inference kernels. We collaborate with open-source toolmakers to bring free cloud compute and venue space to our hackathons.
JOIN 2,500+ BUILDERSGot an idea?
Got an idea?
Come build it with us.
Whether you're fine-tuning models on a gaming laptop at home, building your first agent workflow, or optimizing production inference — there's a seat and a slice of pizza for you.