Invalidity & Prior-Art Search

Prior-art search by a former USPTO patent examiner.

Invalidity searching for litigation and IPR — references located, dated, and mapped element-by-element to the asserted claims, in the form your contentions need them.

Classification-anchored search, not keyword search. Every reference retrieved from a primary source and read against the original document before it reaches you.

Who is doing the search

Most searchers have never examined a patent application.
I spent my career on both sides of the search.

  • USPTO Patent Examiner — internal combustion engines art unit. Searched and applied prior art under examination standards, and know how classification is actually assigned.
  • Patent Engineer & Analyst — Ensemble IP, Dominion Harbor, Concept IP. Patentability, FTO, landscape, and portfolio analysis on both the assertion and defense sides.
  • Mechanical Engineer, NSWC — 8 years at Naval Surface Warfare Center, Department of Defense. Mechanical, electromechanical, and sensing hardware.
  • Lapsed Secret clearance — pre-vetted status available for defense-adjacent matters with a sponsor.
  • B.S. Mechanical Engineering, University of Oklahoma.
  • Search engineer — I built the multi-source search stack described below rather than licensing one, so the method can be tuned per matter.

What you receive

Work product your team can use directly, not a list of hits.

Element-by-element claim charts
Each asserted claim broken into limitations, each limitation mapped to a pinpoint citation — column and line, paragraph, or figure element number.
§102 and §103 reference sets
Anticipation references where they exist; where they don't, a minimal combination with the motivation to combine articulated and sourced, not asserted.
Priority-date verification
Every reference checked against the target's actual priority chain — not the filing date — including provisional support where the chain is contested.
Foreign and non-patent literature
CN, JP, KR, EP and WO art routed to English family members where they exist; journal, standards, and conference literature with translations of the relied-upon passages.
Exclusion screening
Everything cited during prosecution is identified up front, so you learn what is genuinely new art rather than what the examiner already considered.
A written search record
Classes searched, queries run, databases covered, and what was ruled out. If the search comes up short, you can see exactly where it was looked.

Method

Classification first. Keywords second.

Keyword search finds art that uses your vocabulary. The references that invalidate a patent are usually the ones that describe the same structure in different words — a different industry, a different decade, a different language. Examiners find those through classification, and that is where I start.

Classification-anchored recall

CPC and USPC scoped searching, with the class set seeded from the closest art actually found — which corrects for targets that were misclassified during prosecution.

Multiple independent sources

USPTO full text, Google Patents worldwide, and EPO OPS for EP and WO descriptions; plus OpenAlex, Semantic Scholar, Crossref, arXiv and standards bodies for non-patent literature.

Full-text reading, not abstracts

Candidate references are read in full description, because the limitation that matters is rarely in the abstract or the claims.

On AI, plainly

No citation I send you was invented by a language model.

I build software, and automation does real work in my pipeline — running queries across sources in parallel, ranking candidates, pulling full descriptions. It does not decide what is prior art.

I learned why the hard way. When I tested language models on suggesting relevant patent numbers, the overwhelming majority of what came back did not exist — plausible numbers, plausible titles, no such document. So verification against primary sources is built into the pipeline itself: every reference is confirmed to exist, its priority date is pulled from the record, and its relied-upon text is read in the original document before it reaches a chart.

If a reference is in your chart, I have read it. That is the whole standard.

Scope & fees

Fixed fee, quoted before the work starts.

Not hourly. You get a number you can put in a budget, and I carry the risk of the search taking longer than expected.

These sit below what a firm or a search vendor will quote, for a straightforward reason: there is no overhead, no account manager, and no junior handoff. The search is run by the former examiner you are speaking to, which is also why I cap how much I take on at once.

Start here

Preliminary screen

No cost

One patent. I run a scoped search and send you a short written summary: the classes searched, the closest references located, and a straight assessment of whether a full search is likely to be worth commissioning.

If the art looks thin, I will tell you that. You are under no obligation either way, and the screen is yours to keep. I take on a limited number of screens each month, so it is worth asking early if your contentions date is close.

Request a screen →

Focused invalidity search

$2,500 · per patent · fixed

One patent, its independent claims. Suited to early case assessment and to deciding whether a matter is worth fighting.

Full invalidity search

$5,000 · per patent · fixed

All asserted claims, including dependents, with alternative grounds charted where they exist. Exclusion screening against the prosecution record included.

What both include

Both search US and foreign patent literature and non-patent literature, with translations of the passages relied on. Both deliver a petition-ready element-by-element claim chart with verbatim pinpoint citations — in the form invalidity contentions and a petition claim chart require — and a written record of exactly what was searched and what was not.

Pricing assumes a single patent; multi-patent matters are quoted per patent with a volume reduction before any work begins. Rush turnarounds under two weeks carry a surcharge — ask early if your contentions date is close.

If the constraint is bandwidth rather than art, I also take overflow — contentions, charts, invalidity work-up — quoted per matter.

Working together

  • Conflicts cleared first — before any matter is accepted, and never both sides of the same dispute
  • Mutual NDA — signed before confidential material changes hands
  • Search and documentation only — I am a former examiner and analyst, not counsel, and I do not render legal opinions on validity or infringement
  • Work product goes to counsel — delivered to the firm, for the firm to evaluate independently
  • Your materials stay yours — anything you send me that is confidential stays on my systems and is never put through a third-party AI service

The preliminary screen costs you nothing and takes one email.

Send the patent number and the posture of the case. You will get back the classes I searched, the closest art I found, and an honest read on whether it is worth going further.