@GitHub_Daily: A researcher with a paper gets asked whether it can be commercialized — and has to search patents, dig through market reports, and estimate tech readiness on their own, taking days of solo legwork. academic-commercialization-agent hands this off to a team of AI agents: drop in a paper PDF…

Researchers with a paper on their hands get asked whether it can be commercialized — and then end up digging through patents, market reports, and tech-maturity estimates on their own, days of solo legwork. academic-commercialization-agent hands that job to a team of AI agents: feed in a paper PDF or a research direction, and out comes a cited commercialization assessment scorecard. The scoring covers technology readiness, manufacturing readiness, patent strength, market accessibility and evidence confidence; risks and opportunities are listed separately, and every conclusion is tagged with its source. What I like about its approach: code does the retrieval and validation up front, and the six model stages can only reason over that curated set. The weighted total is computed by code too, so the model can’t just make up a number. The author published the experiments as well, failures included. For instance, a lean four-agent variant uses 54.89% fewer median tokens than the default six-agent setup. There are 5 finished sample reports live, one of them in Chinese, viewable right away. Folks working on tech transfer or research administration can start by opening the sample reports to see whether it holds up. — # shuxiachai/academic-commercialization-agent Source: https://github.com/shuxiachai/academic-commercialization-agent # Academic Commercialization Assessment Agent Paper in, cited commercialization triage out. A multi-agent LLM pipeline turns a research paper or topic into a scorecard for technology readiness, manufacturing, patents and market, with citations to registered sources and the risks spelled out. Its experiments are published with their protocols, failures included. Tests (https://github.com/shuxiachai/academic-commercialization-agent/actions/workflows/test.yml) License: MIT Python 3.11 | 3.12 ▶ Read a finished report — no key, no sign-up (https://academic-commercialization-agent.up.railway.app/samples) · Open the application (https://academic-commercialization-agent.up.railway.app) · Run locally · 中文 Demo: a finished CAR-T assessment opened from the history list, its scorecard, risks and opportunities with citations, the cited report, and one saved source excerpt A real run from 2026-08-14 (CAR-T cell therapy for blood cancers, deepseek-chat), opened from a local copy in the current interface and captured 2026-10-23; no model or search request was made. It predates today’s automated checks, which the panel states; it is not a current-model result. ## Try it | Way in | What you need | | |—|—| | Read sample reports | Nothing. Five finished assessments, one in Chinese; opening one makes no model or search request | Open samples (https://academic-commercialization-agent.up.railway.app/samples) | | Run a new analysis online | An operator-issued access code or BYOK: your own LLM API key and model, plus a Tavily or Serper search key. Incurs provider usage | Open the app (https://academic-commercialization-agent.up.railway.app) | | Run it on your machine | Python 3.11 or 3.12, uv (https://docs.astral.sh/uv/) and your own keys | Quick start | The samples are also in examples/: two runs from 2026-10-21, such as Hyperloop, and three from July, such as Solid-state batteries for electric vehicles (frozen 2026-07-19 version, 9673a83 (https://github.com/shuxiachai/academic-commercialization-agent/blob/9673a8346c15f11b989ce8fd2d85be484d1c27b3/examples/solid-state-batteries-ev.md)). The screenshot and samples come from past runs; none of them measures current model accuracy or is an up-to-date market assessment. ## What makes it different - Evidence first, model second. Code retrieves, validates and freezes the sources; six LLM stages then reason only over that registry, and the weighted total is computed by code, not the model. - Measured in the open. Protocols, held-out checks and errata live in the repository. A four-agent variant used 54.89% fewer median tokens than the six-agent default, and the first held-out cohort found TRL scores compressed toward 7, under-scoring mature technologies. See Measured results. - Built for runs that fail halfway. An interrupted run resumes from its last validated checkpoint as a new linked run. Paid requests get receipts, so after a lost reply the result can be recovered for up to 24 hours without paying again. Built with Python, CrewAI, FastAPI and a build-free JavaScript client. Start here: documentation and repository map · Case study Want to contribute? Issues labelled good first issue (https://github.com/shuxiachai/academic-commercialization-agent/labels/good%20first%20issue) need no API keys: the test suite and browser smokes make no paid requests. Start with CONTRIBUTING.md. Finding your way around: the application is src/academic_agent/, api/, web/ and ui/; tests are in tests/ and e2e/. The many Python files at the root are benchmark commands and frozen experiment runners whose protocols bind their paths, so they stay put; none is needed to run the app. See the repository map for every directory. ## What it does - Submit a research topic or attach a paper PDF; choose the report language and scoring profile. - Deterministic code retrieves and validates sources first; six LLM stages then reason over that frozen evidence. Autonomy is limited on purpose. - The scorecard covers TRL, MRL, patent strength, market accessibility and evidence confidence. The weighted total is computed by code, not the model. - Follow progress, inspect citations and reliability warnings, export Markdown or PDF, and share a run link. - In Sources, find a saved source by ID or keyword and read its saved text without a model call. Saved text is not the paper’s full text or verified claim support; see the operating guide. - Optionally add Decision Context: the report distinguishes exploratory orientation from a specific actor’s decision, and whether a threshold was approved by the decision owner. - Resume an interrupted run from its last validated checkpoint. The resumed run is a new linked run; the original is never overwritten. - Paid requests get durable receipts: after a lost reply or a refresh, the accepted run, PDF or resumed run can be recovered for up to 24 hours without paying again. A report supports research triage; it is not technical, legal, regulatory, investment or freedom-to-operate due diligence. A valid citation ID does not establish that the source supports the claim. ## Architecture text Topic / PDF + optional Decision Context │ Deterministic retrieval → validation → frozen source registry │ ┌─────────┼─────────┐ Academic Patent Market └─────────┼─────────┘ Writer │ Reviewer │ Scorer → deterministic weighted total │ Shared run artifacts + final run status │ FastAPI / browser / CLI / recovery The three evidence specialists run in parallel; Writer, Reviewer and Scorer follow in sequence. A stage is not a promise of exactly one model request. See the code map and contribution rules before changing orchestration. | Boundary | Implemented behaviour | |—|—| | Evidence | Source-native clients plus web search; URL/DOI checks, provenance tiers, deduplication and registered source IDs | | Output | Pydantic contracts, guardrails, deterministic scoring and bounded Reviewer corrections | | Quality | Non-blocking, precision-first claim/citation screens; a check that could not run is reported separately from a pass | | Runtime | One subprocess per run, content-addressed checkpoints, recovery into a new linked run, and a final status written exactly once | | Cost | Shared admission for runs and PDF extraction, a persistent daily quota, and usage reported as complete, lower-bound or unavailable | | Observability | Optional redacted OpenTelemetry/OpenInference traces to Phoenix or another OTLP collector | | Delivery | FastAPI, vanilla HTML/CSS/ES modules, Docker and Railway; one application replica | ### Measured results The frozen baseline contains 10 topics × 3 live repetitions: | Check | Observed | |—|—| | End-to-end completion | 30/30 | | TRL calibration | 26/30 | | Weighted formula correctness | 30/30 | | Complete report structure | 30/30 | | Unsupported numeric lines | 0 across 30 reports | These are different checks, not a combined accuracy score. Expected TRL ranges were adjusted after early observations, so this is not independent held-out validation. The uncited-numeric proxy does not measure all hallucinations. Seven of ten topics met their TRL range in all three runs. Held-out check. On ten new topics with ranges fixed from external evidence before any run (qwen3.7-plus, 2026-10-18), 21/30 runs fell in range, the pre-registered “supported” threshold. Scores were compressed toward 7, though: no run scored above 7, so mature, deployed technologies were under-scored. See the held-out result. A second held-out cohort ran on 2026-10-19 with qwen3.8-max and the revised market retrieval. 25/30 runs fell in range, and 11 runs scored 8–9 on proven technologies. Low-maturity topics were not inflated, though the two lowest ones were scored below their floor. The topics and the model also changed, so the improvement is not attributed to any single change. Treat TRL as a coarse signal. See the v2 result. The public deployment rotates through Model Studio’s free-quota models and is not pinned to either cohort’s model; the one it currently requests is in /health as llm_model, and each run page names the model that produced it. No held-out cohort has measured those models. Other completed evidence: - 90-cell topology ablation: the four-node arm used 54.89% fewer median tokens and 47.03% lower median cost than the six-node arm. Six nodes were not established as universally necessary. - Five-reviewer utility study: 20 eligible judgments; the registered success rule failed despite a 6:4 full-workflow preference in each round. - Two target-user pilot: both retained DEFER and answered MAYBE to reuse; neither checked external sources. This does not establish product adoption. - Recovery: 30/30 offline fault-injection recoveries completed; one production recovery reused four committed nodes. This is not an exactly-once or general cost-saving guarantee. - Runtime RTI02: one normal Qwen completion passed 12/12 primary final-status checks, with a disclosed minor observer-cadence deviation. Timeout/fallback paths and general report quality were not validated by that run. Protocols, source artifacts and limitations are linked in the current evidence ledger. ## Quick start Use Python 3.11 or 3.12 for the CI-tested environment and uv (https://docs.astral.sh/uv/). Dependency installation needs network access; the default test suite does not call providers. bash git clone https://github.com/shuxiachai/academic-commercialization-agent.git cd academic-commercialization-agent uv sync Copy .env.example to .env. For Qwen, set these values and replace the placeholders locally; do not commit or share keys: dotenv LLM_PROVIDER=qwen DASHSCOPE_API_KEY=your-key QWEN_MODEL=qwen3.8-27b QWEN_API_BASE=https://dashscope.aliyuncs.com/compatible-mode/v1 TAVILY_API_KEY=your-search-key Choose the endpoint matching your operator account/region. This project’s browser BYOK default uses the China-region endpoint. With multiple LLM keys present, set LLM_PROVIDER explicitly; otherwise auto-selection is DeepSeek → Qwen → Anthropic → OpenAI. Tavily takes precedence over Serper when both search keys are present. Remove unused placeholder keys. bash uv run uvicorn api.main:app --reload # Browser: http://localhost:8000 (sample reports: http://localhost:8000/samples) # Alternative CLI (starts real provider work): uv run academic_agent --topic "solid-state batteries for electric vehicles" Real analysis incurs provider usage. Duration is provider-bound: observed Qwen completions include 306 and 885 seconds, not a promised three-minute SLA. See the operating guide for HTTP endpoints, output files, Docker, access codes, BYOK, tracing and recovery. ## Security and deployment Run URLs carry 128 bits of randomness and act as read capabilities: anyone with the full URL can read that run until retention removes it. Changing a code-owned run additionally requires its owner/admin code; ownerless BYOK runs have no second server-side identity. Do not publish private run URLs. Before public deployment, configure access control, paid-operation limits, retention and persistent storage. PDF extraction is a paid operation too. Use one application replica / one Uvicorn worker: in-memory ownership and file-backed quotas are not a distributed queue. See deployment controls and checkpoint recovery. ## Tests and benchmark The pre-consolidation baseline at 0fdaa76 passed 2071 tests and 678 subtests. CI covers Linux/Windows × Python 3.11/3.12, latest Ruff, narrow Pylint, an 85% coverage floor, zero-provider Chromium smoke, and Docker. These do not constitute a paid-provider production SLO. bash uv run pytest -q uv run --with ruff ruff check . # Scheduling preview only; no provider requests: uv run python benchmark.py --dry-run ### Benchmark New measurements use immutable versioned batches; historical outputs are not overwritten. Reuse requires an explicit matching batch and intact artifacts. Reported costs cover Crew nodes, not helper/PDF/search charges. See the execution and accounting contract. | # | Topic | Expected TRL | Industry | |—|—––|———––|———| | 01 | CAR-T cell therapy for blood cancers | 7–9 | Biomed | | 02 | mRNA vaccines for cancer immunotherapy | 6–8 | Biomed | | 03 | solid-state batteries for electric vehicles | 5–7 | Energy | | 04 | perovskite solar cells for utility-scale power generation | 6–8 | Clean Energy | | 05 | CRISPR gene editing for genetic diseases | 7–9 | Biomed | | 06 | carbon capture and storage for industrial emissions | 6–8 | Climate | | 07 | cultivated meat for food industry | 6–8 | Food | | 08 | quantum computing for drug discovery | 2–4 | Computing | | 09 | graphene-based flexible electronics | 3–5 | Materials | | 10 | room temperature ambient pressure superconductors | 1–2 | Materials | The benchmark guide explains repetitions, frozen evidence and paid-run precautions. Do not change scoring rules to improve these historical numbers. ## Repository layout The application is in a handful of directories. Most of the Python files at the repository root are something else: frozen experiment runners. | Path | What it is | |—|—| | src/academic_agent/ | Pipeline, retrieval, scoring and runtime. Some modules are experiment libraries; the code map says which | | api/, web/, ui/ | FastAPI server, browser client and shared presentation | | tests/, e2e/ | Offline test suite (no provider calls) and opt-in browser journeys | | examples/ | Sample reports, also served at /samples | | scripts/ | Operational report, offline audits and scripted demos | | docs/ | Current guides plus dated protocols, results and errata; start at the map | | Root benchmark*.py | The benchmark commands above | | Other root *.py (openalex_*, evidence_gap_*, report_evidence_*, ablation*, …) | Experiment runners. Not needed to run or understand the application | The experiment runners stay at the root deliberately: their protocols lock source file paths and hashes, so moving them would invalidate those records (see the experiment index). New tools go under scripts/. ## Limitations and open questions Open gaps: qualitative source entailment, independently measured decision utility, multilingual/short/non-technical benchmark coverage, legacy metadata read-state clarity, and single-replica scale. More agents, a vector database or Kubernetes would not by themselves resolve these. Market scoring. Market scorecards disclose [unverified estimate comparabili