Devpost
Active Sprint · Oct 14
Amazon Developer: Building the Future of Cloud & AI
CloudPulse AI
Autonomous AWS Well-Architected & FinOps Governance Agent
Autonomous cloud governance engine powered by Amazon Bedrock Converse API standard. Ingests Terraform/OpenTofu HCL manifests, executes multi-pillar Well-Architected audits (Security, FinOps, Sustainability), projects cost savings, and synthesizes zero-downtime unified Git remediation patches.
7/7 Test Suite Passed: $4,080/yr savings identified · Graviton3 carbon offset
Amazon Bedrock
Claude 3.5 Sonnet
Converse API
TypeScript
Terraform HCL
Devpost
Active Sprint · Oct 18
AWS Cloud Development Series: Agentic AI Edition
DispatchAgent.AI
Autonomous AWS Telephony Voice & Strands A2A Incident Response
Replaces sluggish on-call paging apps with interactive voice response. Outbound calls via AWS Chime SIP Trunking speak directly with engineers, parse voice commands through Amazon Bedrock, coordinate swarms via open AWS Strands A2A protocol, and execute automated ECS rollbacks in seconds.
7/7 Test Suite Passed: Sub-second speech tool calling · Strands A2A protocol
AWS Strands A2A
AWS Chime SIP
Amazon Bedrock
TypeScript
CloudWatch Alarm
Devpost
Active Sprint · Oct 26
Nebius AI Studio & NVIDIA NIM Hackathon
TensorMesh RPA
Ultra-Low-Latency Visual Desktop Automation via Nebius & NVIDIA H100
Autonomous OS-level Robotic Process Automation powered by Nebius Token Factory and NVIDIA NIM on H100 SXM GPU clusters. Delivers sub-50ms TTFT with FP8 TensorRT-LLM quantization. Grounds instructions into screen coordinates and executes zero-drift multi-step desktop actions without APIs.
Dual-Engine H100 Accelerated: Sub-50ms TTFT · 186.5 tps sustained throughput
Nebius Token Factory
NVIDIA NIM
TensorRT-LLM FP8
Llama 3.1 70B
UI Grounding
Lablab.ai
Shipped & Verified
AssemblyAI Voice Agent Hackathon
VoxDive
Real-Time Conversational Voice Agent Over Video & Podcast Knowledge
Unlocks trapped enterprise video training knowledge. Combines AssemblyAI asynchronous STT, speaker diarization, auto-chapters, LeMUR Claude reasoning, and real-time Voice Agent WebSockets. Delivers spoken answers grounded in video with interactive timestamp citation jumping.
Turnaround < 300ms: Live latency HUD · Deterministic offline fallback mode
AssemblyAI Voice Agent
LeMUR Claude
AudioWorklet
WebSockets
TypeScript
Devpost
Shipped & Verified
RevenueCat Shipaton 2026
HabitFlow Pro
Adaptive AI Habit Coach with Native RevenueCat Subscriptions
Cross-platform mobile application built on Expo & React Native with an adaptive rule-based habit intelligence engine, gated behind RevenueCat `pro_access` entitlements. Features Paywall packages (Monthly, Annual, Lifetime), restore purchases, and offline persistence.
36/36 Unit Tests Passing: Simulated store sandbox · Zero runtime crashes
Expo / React Native
RevenueCat
react-native-purchases
TypeScript
AsyncStorage
Devpost
Ready to Submit
OpenCV AI Competition 2026 (Powered by AWS)
EdgeSentry
Edge Computer Vision Workplace Safety & Restricted Hazard Monitor
Industrial camera monitoring system built with genuine OpenCV geometric primitives. Runs edge ONNX object detection, pointPolygonTest spatial hazard boundary checks, HSV hi-vis PPE inspection, and dispatches verified incident snapshots to AWS S3 & SNS alarms.
Real OpenCV Pipeline: Temporal persistence debounce · Multi-zone hazard assertion
Python 3.10+
OpenCV cv2.dnn
AWS SNS / S3
ONNX / YOLO
Spatial Geometry
DoraHacks
Ready to Submit
Galuxium Nexus V2 Hackathon
NexusGraph Enterprise
Hybrid GraphRAG Enterprise Knowledge Mesh & Root-Cause Lineage
Unifies fragmented corporate knowledge (Jira tickets, GitHub PRs, Datadog APM traces, and Workday org graphs) into a typed knowledge graph. Fuses dense vector similarity with multi-hop topological graph traversal using Weighted Reciprocal Rank Fusion (WRRF k=60) to eliminate hallucinations.
5/5 Test Suite Passed: 3-hop causal root-cause · 100% provenance verification
Hybrid GraphRAG
WRRF Fusion (k=60)
Multi-Hop Traversal
TypeScript
Enterprise Mesh
HackerEarth
Shipped & Benchmark Passed
micro1 First AI Hackathon
Triage Inbox
Evidence-First Multi-Agent Maintainer Copilot for GitHub PRs & Releases
Audits software release notes and pull request reviews using a multi-agent evidence-first architecture. Eliminates naive LLM false alarms by grounding every single claim in physical Git commits, diffs, and AST changes before presenting findings to the human maintainer.
Precision 1.00 vs 0.70 Baseline: Zero false alarms · Mean F1: 0.87 (0.82–0.95)
Python
Multi-Agent Swarm
Git Diff Grounding
OpenAI / Claude
Human-in-Loop
Kaggle
Competition Benchmark
ARC Prize 2026 Code Competition
Neuro-Symbolic ARC Solver
Symbolic Program-Synthesis DSL Engine for ARC-AGI Abstraction Tasks
Symbolic program-synthesis solver for ARC-AGI grid-reasoning tasks. Searches compositions of pure NumPy DSL primitives (rotations, reflections, transpose, gravity, tiling, and color maps) to synthesize programs reproducing demonstration pairs with exact-match verification.
91.7% Solve Rate (88/96 tasks): Pure NumPy DSL · Zero GPU dependency
Python 3.10+
NumPy DSL
Program Synthesis
Kaggle submission.json
Kaggle
Competition PoC
CZ Biohub — Cell Tracking During Development
3D Spatiotemporal Cell Tracker
Kalman-Assisted Hungarian Association for 3D Microscopy Lineage
Tracks dividing cells across 3D developmental microscopy time-lapses. Employs constant-velocity anisotropic Kalman filters, Hungarian optimal data association (`scipy.optimize.linear_sum_assignment`), mitosis identification, and Cell Tracking Challenge (CTC) format exports.
Purity 1.0 · 0 ID Switches: Optimal Hungarian association · CTC res_track.txt
Python
SciPy Hungarian
Kalman Filter
3D Connected-Component
CTC Format
Kaggle
Simulation Agent
Kaggriculture Simulation Competition
Kaggriculture MCTS
UCT Monte-Carlo Tree Search Planning for Digital-Twin Farming
Autonomous agricultural agent maximizing seasonal yield and revenue under stochastic weather and soil moisture dynamics. Features a cloneable digital-twin environment model paired with an Upper Confidence bounds for Trees (UCT) Monte-Carlo Tree Search policy.
UCT Tree Search Engine: Time-capped 40-iteration search · Sustainability scoring
Python
MCTS UCT
Forward Simulation
Greedy Rollout
Digital Twin
Kaggle
Competition Model
RSNA Knee Abnormality Detection
RSNA Knee Abnormality Detection
Multi-Sequence 2.5D Volumetric MRI Classifier & QWK Metric Calibration
Multi-sequence knee-MRI abnormality classifier assessing ACL tears, meniscus injuries, and cartilage defects. Incorporates volumetric NumPy slice standardization, percentile z-score normalization, multi-view 2.5D ResNet-18 fusion, and Quadratic Weighted Kappa metric scoring.
QWK Calibrated Metric: 2.5D Sagittal/Coronal/Axial fusion · Kaggle submission.csv
Python
NumPy Volumetric
PyTorch ResNet
Weighted Kappa
DICOM Processing
Lablab.ai
Shipped & Verified
AMD Developer Hackathon: ACT III & AI Academy
OmniMesh-ROCm
Enterprise Multi-Agent MCP Mesh on AMD Instinct MI300X
High-performance Model Context Protocol (MCP) multi-agent orchestration engine optimized for AMD Instinct MI300X GPU accelerators and ROCm 6.2. Enables autonomous subagent swarms to dynamically register MCP tools, stream FP8 vLLM-ROCm inference, and verify AST execution traces.
284.5 tok/s (FP8 MI300X): 38.2ms TTFT · 100% AST trace grounding
AMD ROCm 6.2
Instinct MI300X
MCP 2026
vLLM FP8
TypeScript
Lablab.ai
Shipped & Verified
Vultr: Agent Rush Hackathon
VultrSwarm-Agentic
Autonomous CloudOps & Multi-Agent Infrastructure Swarm
Autonomous multi-agent cloud infrastructure orchestrator that continuously monitors Vultr compute clusters, detects CPU/memory saturation anomalies, and autonomously dispatches serverless worker containers and vertical instance scaling without human intervention.
3/3 Test Suite Passed: Zero-touch Vultr scaling · Immutable CloudOps audit log
Vultr Cloud API
Multi-Agent Swarm
CloudOps
Auto-Remediation
TypeScript
NeurIPS 2026
Active Sprint · Nov 21
IBM Research & NeurIPS 2026 Steerability Challenge
HonestSteer
Standardized .spipe Activation Steering Pipeline for LLM Truthfulness
Implements a standardized .spipe residual-stream activation steering pipeline that injects calibrated honesty direction vectors during forward passes to suppress dishonest assertions without degrading MMLU/GSM8K reasoning.
2/2 Test Suite Passed: Standardized .spipe export · Zero capability collapse
Activation Steering
.spipe Standard
NeurIPS 2026
IBM Research
Python 3.11
NeurIPS 2026
Active Sprint · NeurIPS Track
NeurIPS 2026 Agenthon (SQA & CEWIT Stony Brook)
VeriQuant AI
Leak-Proof Point-in-Time Quantitative Alpha & Portfolio Agent
Autonomous quantitative alpha generation and dollar-neutral mean-variance portfolio optimization agent designed to pass strict leakage-controlled, cheat-resistant G0–G3 evaluation gates.
G0–G3 Gates Certified: 3/3 Tests Passed · Strictly lagged (t-1) zero lookahead
Quant Alpha
Mean-Variance Opt
G0-G3 Audit Gates
Python 3.11
Docker
Lablab.ai
Active Sprint · Nov 3
AI Genesis 2026 Global Hackathon (Lablab.ai)
Genesis Ops Mesh
Planner-Executor-Verifier Multi-Agent Enterprise Workflow Engine
Autonomous multi-agent enterprise operational workflow orchestrator combining structured DAG task decomposition, parallel tool execution swarms, and deterministic ground-truth verification gates.
100% Deterministic Gate: Strict TypeScript · Zero unverified tool outputs
TypeScript (Strict)
Multi-Agent Swarm
Verifier Gate
Lablab.ai
No hackathon entries match your filter
Try clearing your search query or selecting a different platform tab.