Devpost
Participate in our public hackathons
Devpost for Teams
Access your company's private hackathons
Grow your developer ecosystem and promote your platform
Drive innovation, collaboration, and retention within your organization
By use case
Blog
Insights into hackathon planning and participation
Customer stories
Inspiration from peers and other industry leaders
Planning guides
Best practices for planning online and in-person hackathons
Webinars & events
Upcoming events and on-demand recordings
Help desk
Common questions and support documentation
A reproducible parameter ensemble for AI-image detection that stays robust under JPEG, blur, resize, noise, color jitter, and crop.
Tiktalk: An AI shopping assistant that understands intent, remembers context, adapts to changing preferences to deliver smarter, more personalized product recommendations through conversation
The Renegaders remembers what shoppers mean, asks one useful follow-up, and finds the right product from 50,000 options in 93.5% of released test sessions, with zero model cost.
Finds the exact product a shopper means in 1.85 turns on average. Pure rules + BM25 — no LLM, no API, zero tokens, fully offline. TechnicalScore 0.912 against a 0.107 baseline.
A deterministic, mathematically-bounded e-commerce agent that eliminates LLM hallucination and slashes Mean Turns to Completion (MTTC) by 25%
A conversational shopping copilot that remembers every requirement you mention, then retrieves and re-ranks in real time, surfacing the right product within the top 10 in roughly two turns.
AI generated content detector
TAISP is back with a brand new invention
EpsilonRec is not one recommender model, but an autonomous ML researcher that hypothesizes, trains, evaluates, and revises KuaiRand-Pure recommenders with zero human intervention.
An autonomous agent that does ML research: hypothesise, code, run, measure, repeat. It beat the official KuaiRand recommender baseline in 30 self-directed iterations with zero human intervention.
Agent middleware that runs a plan's tool calls before the model has finished writing the plan, and refuses the ones it cannot prove safe. 2.19x faster end to end
A conversational shopping agent that remembers what users want, reject, and change—then combines lexical and semantic search to find the right product within ten turns.
A Transformer layer made 6.54x faster with every optimization measured and every dead end recorded
AI agents and custom GPU kernels accelerate Transformer inference by 3.9x on an RTX 4070 with bitwise-identical outputs.
Every shape gets its own kernel, driven by research agents and heuristic tuning policies.
An autonomous ML research agent that experiments, learns from failures, and improves recommender systems with minimal human intervention.
An autonomous agent that improves recommendation models through its own experiment loop
Fast Retail is a blazing-fast, 100% local e-commerce agent. It uses SQLite and BM25 to deliver instant product recommendations without the cost, latency, or privacy risks of LLMs.
A robust AI-image detector built for the real world with a 0.9867 robustness score and 0.9986 AUROC on unseen generators after screenshots, compression, resizing, and cropping.
Zero-trust execution middleware where AI agents can request authority—but never grant it to themselves.
A multi-turn AI shopping copilot that understands evolving user needs and combines hybrid retrieval, intelligent clarification, and evidence-aware ranking to find the right product faster.
GoalThread turns fragmented AI agent tasks into goal-aware threads! Merging related work, isolating unrelated tasks, and forking cleanly when your goal shifts.
An autonomous ML research agent that proposes hypotheses, edits recommender code, trains and evaluates candidates, and learns from signed results with zero manual intervention.
Give every AI agent its own identity and enforce owner-controlled, auditable and revocable access at runtime, including secure cross-agent data sharing.
73 – 96 of 614
