Senior Frontend Engineer Mumbai · 4+ years · Full-time →

NyaayOS; Mumbai; On-site; INR 1,800,000 - 2,800,000; Full time

Job summary

ParamAIQ in Mumbai is seeking a senior frontend engineer to own the architecture and code powering a legal research platform. You will drive the frontend for a streaming research assistant, live citations, and a drafting editor with citation-aware objects.

The role emphasizes building accessible, robust UI with end-to-end tests, handling long documents, and maintaining performance on mid-range laptops and variable network conditions.

Qualifications

• Shipped a complex web app in TypeScript and React or similar.
• Streaming and async UI: server-sent events or WebSockets, cancellation, and the error and partial states most apps skip.
• Accessibility and tests are part of done for you, not a later ticket.

Responsibilities

• Own the architecture and the code of the app lawyers and firms log into.
• Implement streaming progress, live citations, and correct source linking for each claim.
• Ensure accessibility, robust tests, and reliable error handling across the frontend.

Job description

A lawyer won't file anything on the strength of an answer they can't check. So in ParamAIQ, the interface is where trust is won or lost. Every answer has to show where it came from, how sure we are, and whether the law it cites is still good, and the lawyer has to be able to verify it in one click. You build that: search, a research assistant that streams its work as it goes, a drafting editor where citations are live objects, and a document viewer that opens a 300 page judgment at the exact paragraph. This is a senior, hands-on role. You make the frontend architecture calls, and you write the code that matters most.

YOU WILL OWN

• The architecture and the code of the app lawyers and firms log into.
• Search over millions of judgments, orders, and statutes. Filters by court, judge, section, and date, and results a lawyer can scan in seconds.
• The research assistant. Answers that stream in, live progress while it searches and reads, and a citation on each claim that opens the source at the right paragraph.
• Evidence on every AI answer. Its sources, how confident we are, and whether a cited judgment has since been overruled.
• The drafting editor. Citations and defined terms as live objects, formatting that follows each court's rules, and PDF and Word export that's ready to file.
• The document viewer. Long judgments in PDF or text, with the cited passage highlighted.
• Matter workspaces, where a firm keeps its research and drafts, visible only to the people who should see them.
• Speed in real conditions: long documents, streamed answers, mid-range laptops, and office connections that drop.
• The design system, built with our designer, so the product looks and behaves as one thing.
• The internal tools the rest of the company runs on. Our legal, research, and operations teams need software too, and it gets the same care as the product.

THE HARD PROBLEMS

• Showing uncertainty without hiding the answer. An AI answer can be part grounded, part inferred, and occasionally lean on a judgment that has since been overruled. Making that visible at the level of a single claim, without burying the lawyer in warnings, is not a solved problem in any product we know of.
• Making slow, fallible AI feel dependable. Answers arrive over seconds. Steps run, fail, and retry. The interface has to show progress honestly, handle partial and failed results gracefully, and never present a half-finished answer as a finished one.
• An editor that knows what it contains. A legal draft isn't rich text. Clauses, citations, and defined terms are structured objects that can be checked, and flagged when the law under them changes. The lawyer still has to type as freely as they do in Word.

WE ARE LOOKING FOR

• You've shipped a complex web app that people use for real work, in TypeScript and React or a similar framework, and you can talk through a hard architecture call you made in it.
• Streaming and async UI: server-sent events or WebSockets, cancellation, and the error and partial states most apps skip.
• Accessibility and tests are part of done for you, not a later ticket.

NICE TO HAVE

• A rich-text editor built on ProseMirror, Lexical, or similar. Taken well past the default setup.
• AI products. Streamed model output, citations, confidence, or human-review interfaces.
• Document-heavy interfaces. PDF rendering, virtualised long documents, highlights and annotations.
• A design system built from zero.

THIS ISN'T THE ROLE IF

• You think a spinner is a finished design for a ten second wait. Most of the hard work is in the states between "loading" and "done."
• You want a mature design system to work inside. We're building it.
• You want the backend to hand you finished data. Deciding with backend and research what an answer must carry, so the screen can show it honestly, is part of the job.
• You want to architect and hand off. You'll be in the code, down to the empty states.