November 20, 2026

Garage Inference Hackathon 2026: 37 Teams Built AI Tools on Models That Run on a Laptop — Then Engineers from Sonos, Amazon, and Gorgias Asked Whether Any of It Would Survive Production

real_c7016298-ad84-426c-bac7-3598b765a173
Industry veterans in data engineering, cloud infrastructure, and applied AI assessed whether tools built on sub-billion-parameter “garage” models hold up beyond the demo.

Hackathon Raptors, a UK Community Interest Company (CIC #15557917), has concluded Garage Inference Hackathon 2026, a competition built around a deliberately awkward constraint: build something genuinely useful using only “garage-grade” language models — models small enough to run on a laptop, frequently under a billion parameters, where raw capability is scarce and every other engineering decision has to compensate. Across a 72-hour build window in early May, 37 teams shipped tools that, by the measure of model size alone, had no business working as well as they did. The tagline the organizers chose — “Big Ideas, Cheap Models. The constraint is the creativity.” — turned out to be an accurate description of what the judging surfaced.

The premise inverts the prevailing direction of the AI industry, which has spent three years racing toward larger models and bigger context windows. Garage Inference asked the opposite question: when the model is the cheap, weak part of the system, what does good engineering actually look like? The answer, according to the panel that evaluated the work, lives almost entirely in the scaffolding around the model — the data discipline, the deployment posture, the verification layers, and the operational maturity that separate a tool from a tech demo.

How the Projects Were Judged

Submissions were scored on a six-criterion framework weighted toward the event’s central question. “Results vs. Model Capability,” the so-called Wow Gap measuring how far a system exceeded what its underlying model could do alone, carried the most weight. Practical Usefulness and Technical Execution followed, with Creativity and Innovation, Accessibility and Reproducibility, and Secure Design rounding out the rubric. Final standings were computed with a weighted average and a Bayesian adjustment to equalize for the fact that different projects drew different numbers of judges — a methodological step intended to keep a project reviewed by four judges comparable to one reviewed by eight.

The evaluation panel was drawn from senior engineers across consumer hardware, cloud infrastructure, and AI product organizations.

Manushi Sheth, who leads the Product Data team at Sonos across data engineering, analytics, and machine learning for millions of connected devices, evaluated projects through the lens of data discipline. Her recurring concern was that small models fail differently from large ones: “A small model is a magnifying glass held over your data. Whatever discipline you skipped, it finds and amplifies.” She flagged the category’s defining risk as silent failure — a model with no capacity to signal its own uncertainty narrating a wrong answer as fluently as a right one — and rewarded the teams that built explicit verification boundaries between the model and anything that trusted its output.

Sumanth Kadulla, a cloud infrastructure and DevOps engineer with more than eight years building high-availability systems across AWS, Azure, and Google Cloud and a technical reviewer for over twenty-five international conferences, assessed projects on deployability. His standard was unsentimental: “Cheap inference is only cheap if you can actually deploy it cheaply. A model that runs locally is a starting point, not a finish line.” He reserved his highest marks for projects that treated operations — cost per request, cold-start behavior, failure isolation, secure handling of inputs — as first-class concerns rather than afterthoughts.

Deniz Aleyna Akbasaran, a product and data specialist on the Product and Data team at Gorgias in New York, where she builds evaluation frameworks for AI agents and leads cost-optimization for large-scale AI systems, brought an evaluation-first perspective sharpened by a Columbia master’s in management science and engineering completed as a Fulbright Scholar. Her attention centered on whether teams could demonstrate that their tools actually worked, rather than merely appeared to: “The hard part isn’t building the agent. It’s proving it does what you claim, repeatably, on inputs you didn’t choose.”

Nishant Sinha, an engineering lead with nine years building distributed and machine-learning systems and deploying models at the edge, evaluated the projects against the realities of running inference outside the data center. His framing drew on hard-won experience: “Deploying a model at the edge teaches you fast that the model is the easy part. The system around it — the part that handles the device you didn’t test on — is the whole job.”

What the Teams Built

The top of the field was defined by projects that used deterministic engineering to lift a weak model into genuinely reliable territory. EDGEDOCTOR AI, from team BharatEdge, took first place and the event’s Root Access Award: an offline medical triage system that wrapped a 0.6-billion-parameter model in seven layers of deterministic engineering, including a real retrieval system and a substantial test suite. Where the raw model, given “chest pain, left arm numb,” responded with “stay hydrated and rest,” the engineered system produced safe triage — a vivid demonstration of the Wow Gap the event was built to measure.

Atomic PR Surgeon, from team RAG Tag, finished second: a zero-cost code reviewer that ran an ensemble of four specialized micro-agents on a 600-million-parameter model to catch SQL injection, N+1 queries, and cross-file vulnerabilities, then generate fix patches. TinyMind_Coder, from TinyMind Labs, took third with an approach centered on structured reasoning loops, execution feedback, and lightweight verification — pushing the limits of small models through inference engineering rather than scale. A Spotlight award for the widest Wow Gap went to Marionette, a browser-based assistant that ran entirely on the user’s own machine and answered only to its owner.

Across the field, the panel kept returning to a single pattern: the projects that succeeded treated the model as one untrusted component inside a larger, mostly deterministic system. Verification layers, caching, lineage, and local-first deployment recurred far more often among the strongest entries than any particular modeling trick.

The judging also surfaced an uncomfortable lesson about honesty under constraint. Because the projects were small and inspectable, the panel could read the code rather than take the demo on faith — and a working demo was, in several cases, no guarantee that the advertised features were real. The teams that fared best were those whose claims survived inspection: a retrieval system that actually retrieved, a verification layer that actually verified, a deployment story that actually deployed. For judges drawn from production engineering organizations, that emphasis on verifiable substance over polished presentation was not incidental; it mirrors the standard their day jobs demand, where a feature that works in a demo and fails under load is worse than no feature at all.

The Pattern Beneath the Results

For the organizers, the value of the constraint was precisely that it accelerated lessons that usually take a production incident to learn. By removing the cushion that a frontier model provides, Garage Inference forced teams to confront — in a weekend — the data debt, deployment cost, and verification gaps that more typically surface in month two of a product’s life. The judges’ consensus was that these lessons travel upward: the discipline that makes a 600-million-parameter model reliable is the same discipline that makes any AI system reliable, simply made unavoidable by the constraint. It is also a discipline the wider industry is moving toward for reasons of cost, latency, and privacy, as a growing share of production AI shifts to smaller models running closer to the user. The engineers who learned to make a weak model trustworthy under constraint, the panel suggested, were rehearsing the exact skill the field is about to need at scale — and the strongest entries read less like weekend experiments than like early drafts of how serious small-model software will be built.

Hackathon Raptors, a UK Community Interest Company (CIC #15557917), is a London-based organization that conducts technology challenges testing engineering fundamentals under constraint. Its fellowship model brings together engineers from organizations including Sonos, Amazon, and Gorgias to evaluate emerging work. More at raptors.dev or hello@raptors.dev.