Legal practice
His work at A. J. Antony Associates spans litigation, arbitration and commercial matters. This helps him interrogate contractual provisions, the parties’ documentary accounts and the strength of a legal proposition.
Practitioner Evaluation & Collaboration
advocate · ENGINEER · Legal AI builder
Abhishek Antony
Junior Partner at A. J. Antony Associates
Legal practice and engineering expertise brought together to challenge AI-generated legal analysis against the original documents and the human review workflow.
Abhishek provides hands-on practitioner evaluation of LegalRAG Pro, using fictional legal matters to test whether answers are complete, accurately sourced, consistent and genuinely useful for professional review.
Professional background
Experience spanning law, engineering and legal technology
Abhishek is an Indian qualified advocate based in London. He is a Junior Partner at A. J. Antony Associates, where he handles litigation, arbitration and commercial matters and led the firm’s digital transformation. He holds an LLM in Innovation, Technology and Law from the University of Edinburgh and a B.Tech in Mechanical Engineering. He builds legal AI tools in Python and holds a Legal Engineer certification from Harvey. This background is useful when assessing both the substance of an AI-generated legal answer and the software workflow used to produce it.
Engineering and technology
Alongside his mechanical engineering qualification, he has led digital transformation work at the firm and develops legal AI tools in Python. He can therefore assess both legal reasoning and practical software behaviour.
Specialist study
His LLM in Innovation, Technology and Law from the University of Edinburgh and Legal Engineer certification from Harvey connect his legal practice to questions about AI-assisted professional workflows.
Evaluation in practice
What his testing examines
LegalRAG Pro is intended to keep a legal answer connected to its evidence. Abhishek’s evaluations test whether that connection withstands detailed scrutiny, not merely whether the wording sounds convincing.
1. Evidence retrieval and coverage
Using a synthetic documentary dispute containing 35 source documents, he has tested whether differently phrased legal questions retrieve relevant evidence, whether significant materials are missed and whether the response covers each part of the question.
2. Citation integrity and legal precision
He checks legal propositions against original source documents and individual clauses. His reviews have identified where an answer needed a more precise contractual pinpoint and where a statement went further than the document cited in its support.
3. Comparative reasoning
He tests whether the analysis distinguishes a party’s allegation from an independent reviewer’s assessment, separates the causes of delay from the grounds advanced for termination and recognises evidence that qualifies an apparently straightforward conclusion.
4. Consistency across questions
He compares responses to repeated and rephrased questions. That work considers whether the same question produces materially different answers, whether semantic matching finds equivalent terminology and whether conclusions remain grounded as the wording changes.
5. Supervisor review and usability
His feedback covers how questions, investigations and reviewed answers are presented, including the clarity of review headings, the visibility of internal technical material and how easily a supervisor can inspect the basis of the work.
6. Continuity and reliability
He has also reported practical difficulties such as interrupted conversation history, page-loading delays and the need for reliable saved investigations. Such observations matter when legal work needs to be reopened, checked and continued across sessions.
An example from a fictional case
Testing what a source actually says
In a subsequent review of the fictional Aurelian Commodities v Brightwater Systems matter, Abhishek challenged an answer which attributed a connection between alleged staffing shortfalls and project delay to a notice of default. On inspection, that notice did not itself support the proposition as written.
He distinguished it from a separate independent project health check addressing the causes of delay, while also identifying that the alleged contractual grounds for termination needed to be analysed on their own terms. This is the type of distinction a source-linked legal tool must preserve: a plausible narrative is not enough if the cited document does not substantiate it.
From feedback to development
How his contribution informs the product
Run questions and human-review workflows against synthetic evidence.
Compare assertions, citations and omissions with the original material.
Record reproducible examples of answer-quality and usability problems.
Help assess later changes as the engineering team investigates and corrects verified issues.
His combination of legal and technical knowledge makes the feedback useful beyond the content of a single answer. It can inform retrieval, citation verification, governance, saved work and the overall process of professional supervision.
Continuing collaboration
Abhishek has expressed interest in continuing to test LegalRAG Pro and exploring how his legal and engineering background might contribute to future development. Discussions about a potential advisory or development role remain separate from the practitioner evaluation described here; no formal role or partnership has been agreed.
Professional boundaries
Abhishek has consented to the publication of his profile and description of his testing. His involvement is not an endorsement by A. J. Antony Associates, a UK solicitor approval, a regulatory certification or a warranty about the platform. Testing uses synthetic material. Legal professionals remain responsible for verifying original evidence and exercising their own judgement.