Proof
Every accuracy claim, every product, every credential on this site is traceable to a public artifact someone other than me has signed off on. This page collects them in one place so you can verify before a single email is exchanged.
Last updated 2026-04-28.
Public benchmarks
SureCiteAI's accuracy claims are graded against public, peer-reviewed corpora that anyone with the open-source repo can re-run end-to-end. The most recent published run (2026-04-27, Cohere rerank-3.5) scored 221/297 (74%) aggregate with 0/297 hallucinated citations across six suites — including the 150-question PatronusAI FinanceBench external benchmark (NeurIPS 2023, arXiv:2311.11944), where retrieval hit-rate was 96% (144/150) and citation hallucinations were zero.
Methodology, run history, per-suite scorecard, failure-mode breakdown, and reproducibility instructions are public. Anyone with a Pinecone, Supabase, Cohere, and OpenAI account can produce the same numbers.
- SureCiteAI case study (full per-suite breakdown)
- CUAD v1 — Hendrycks et al., NeurIPS 2021 (legal contract benchmark, CC BY 4.0)
- PatronusAI FinanceBench dataset (HuggingFace)
Live products you can use right now
Two production SaaS products built and shipped as sole architect. Both are publicly accessible — sign up, log in, run the product. No demo videos in place of a live URL.
- sureciteai.com — multi-tenant document intelligence RAG SaaS
- systemaudit.dev — automated codebase intelligence reports in under 3 minutes
Open-source code
The benchmark harness, evaluation scripts, and product source for SureCiteAI are public. Commit history, file diffs, test runs, and the raw benchmark artifacts are auditable on GitHub.
- github.com/nicuk — public profile
- SureCiteAI case study — links to
BENCHMARKS.md,benchmarks/runs/, and the evaluation scripts - SystemAudit case study — open-source scanner engine (MIT)
Funded venture
Previously co-founded SculptAI (a 4-agent game development pipeline, 2024–2025) which raised $350K in seed funding. Listed on Crunchbase under the parent company Definity Legend.
Third-party identity
Knowledge-graph identity is registered on Wikidata. This is the same identity layer large language models use when grounding entity claims, and it disambiguates "Nic Chin" from any other person sharing the name.
Independent client reviews
Client engagements run under NDA, so individual client names, contracts, and outcomes are not published. Independent platform-verified review history exists and is shared on request during qualification calls — Upwork Top Rated Plus (top 3% of talent globally), 100% Job Success Score, multiple verified completed contracts.
The live profile link is held back deliberately during a current review-cycle reset; it will be published here once the cycle settles. Verification on request — ask in the first call.
Credentials
Vendor-issued credentials from IBM (AI Agents & RAG), Microsoft (Generative AI), and Google (Prompting), plus a University of Northampton degree.
Full credential identifiers (with the certificate-holder's legal name) are available under NDA on request — published certificate URLs leak the holder's full legal name, which we hold back on this site for personal-security reasons. Issuer is fully verifiable; identifier is shared during qualification.
What is deliberately not on this page
- Named client testimonials with company logos. Engagements run under NDA.
- Direct screenshots of internal client systems, dashboards, or proprietary data.
- Full certificate URLs (which expose the holder's legal name). Issuer is shown; identifier is shared on request.
If something here matters to your evaluation and you can't verify it from this page, ask in the first call.
Want the full technical proof?
The SureCiteAI case study has the full per-suite benchmark breakdown, the citation verifier design, and links to the raw run artifacts.