Evals before prompts: building an LLM OCR for KYC
The article describes how the author built an LLM-based OCR system to extract data from Indian KYC documents such as PAN, Aadhaar, driver’s licenses, and vehicle registrations. A curated “golden dataset” of 50 verified examples per document type was created to serve as a benchmark for evaluating model performance. A detailed rubric and human review process were used to assess accuracy and avoid harmful hallucinations.
- ▪The author assembled a small but representative golden dataset, splitting it into four slices for PAN, Aadhaar, driver’s licenses, and vehicle registration certificates.
- ▪Each slice contained 50 documents, allowing rapid iteration and expert review of model outputs.
- ▪A field‑by‑field rubric defined exact‑match criteria for IDs, name variations, and date formats, and penalized hallucinated fields more heavily than missing data.
- ▪Human reviewers compared LLM extractions to the original scans to catch errors that automated checks miss, especially when the model produced confident but incorrect answers.
Hacker News (AI / LLM) files mainly under ai. We currently carry 2,748 of its stories.
Story provenance
Source · retrieval · rights · ranking — open for full record
inspect →
Story provenance
Attribution is not the same as permission. This drawer separates discovery metadata, excerpts, WeSearch-generated summaries, reuse status, and whether the publisher receives the visit. Nothing here claims a legal grant the publisher has not made.
Record
| Original publisher | nilenso blog |
| Canonical URL | https://blog.nilenso.com/blog/2026/05/18/evals-before-prompts-building-an-llm-ocr-for-kyc/ |
| Publication time | Tue, 28 Jul 2026 06:58:01 +0000 |
| Retrieval time | 2026-07-28T07:14:57.182Z |
| Last seen | 2026-07-28T07:14:57.182Z |
| Headline source | Publisher (no WeSearch rewrite) |
| Excerpt source | publisher body |
| Excerpt method | First ~120 words (~800 chars) of extracted publisher body, fair-use limited. |
| Summary | WeSearch · cerebras-chat (WeSearch summarizer) |
| Summary source text | contentText |
| Citation coverage | Summary is a WeSearch-generated derivative; primary citation is the original publisher URL. |
| Cluster | 8inL-Gg02zdU · 1 stories |
| Cluster logic | Grouped by semantic title/content similarity across sources within a rolling window. Same-publisher template collisions are excluded from coverage comparison. |
| Ranking reason | Story pages are not engagement-ranked. Hub feeds use recency, with optional source-diversified chronological ordering (cap consecutive stories per source). No personalized ranking. |
| Publisher visit | Yes — open original |
| Substitutes article? | No — link-out required for full text |
Rights status (four layers)
WeSearch handling by dimension
| Indexing | May the item be indexed (stored, ranked, made findable)? | Allowed |
| Snippet | May a short excerpt of the publisher's text be shown? | Allowed |
| AI summary | May WeSearch generate its own short summary of the article? | Limited |
| Retrieval / RAG | May the content be exposed for third-party retrieval-augmented generation? | Not asserted |
| Model training | May the content be used to train AI models? | Not asserted |
| Commercial reuse | May the content be reused commercially? | Not permitted |
Basis: Derived from the published RSS/Atom feed. Contact: [email protected]. Reviewed: 2026-07-24.
Opening excerpt (first ~120 words) tap to expand
Priyanga P Kini Evals before prompts: building an LLM OCR for KYC 18 May 2026 KYC, or Know Your Customer, is how banks, fintechs, and a growing number of businesses verify that a person is who they claim to be. At the heart of it sits a deeply unglamorous task: typing the fields off a scanned PAN card or Aadhaar into a form. Humans are bad at this. It’s boring, error-prone, and, at scale, expensive. Can an LLM do it instead? Around this time last year, I was trying to answer that question for one of our clients. They were paying a third-party service to do this extraction for them, and the bill was steep enough that they wanted to bring the work in-house. The question was whether LLMs could reduce the cost of document verification without sacrificing the accuracy KYC demands.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at nilenso blog.
Discussion
0 commentsMore from nilenso blog
