Remote (India)
Adalat AI is building an end-to-end justice tech stack that automates manual and clerical pain points in courtrooms, giving judges back time to focus on what matters most: decision-making and delivering justice. Our solutions - from AI-powered transcription in Indian languages to case-flow management and document navigation - are now deployed across 9 states, covering nearly 20% of India’s judiciary. Backed by leading technology companies and funders, and incubated at MIT and Oxford, Adalat AI is working to eliminate judicial delays and expand access to timely justice. Founded by a team with backgrounds in law, technology, and economics from Harvard, Oxford, MIT, and IIIT Hyderabad, we are scaling rapidly across India and the Global South.
Role Overview
Full-time · Remote (India)
Hindi is the primary working language of courts across North and Central India. Court dictation in these regions reflects significant linguistic variation: regional speech patterns (such as Malvi, Bundeli, Bhojpuri and Rajasthani), frequent Hindi–English code-switching, spoken section numbers and case citations, and legal terminology with multiple accepted spellings. Existing public datasets do not adequately represent this variation.
As the Hindi Language Expert, you will lead the linguistic quality of our Hindi speech and text data. You will distinguish regional pronunciation from transcription errors, define standards for how Hindi legal speech is written, and ensure that the data used to train and evaluate our models is accurate and consistent. As part of the Data team, you will work closely with ML engineers, legal experts and our panel of language annotators.
Key Responsibilities
1. Speech data quality
Write dictation scripts from real court orders that cover general and legal language, and be the final reviewer for the recordings that come back: usable, fix, or re-record.
Turn Hindi dictation into gold-standard transcripts, catching the errors that matter in court: a dropped नहीं, a wrong section number, a misheard name.
2. Text and documents: extraction and translation
Review the quality of Hindi text extracted from court documents, including OCR.
Translate Hindi ↔ English legal and general text, and verify translations produced by our models or open source datasets for accuracy, legal meaning and consistent terminology.
3. Hindi language standards
Decide how Hindi legal language is written: spelling variants (दंड / दण्ड), nukta (ज़मानत / जमानत), abbreviations (अ.सा., भा.दं.सं.), numbers, dates and English legal terms, and keep the style guide and legal glossary that dictation, text generation and translation follow.
Run district-level surveys and interviews with annotators to document regional pronunciation and vocabulary, and map which varieties our data covers.
4. Evaluation and model feedback
Help build the Hindi test sets every new model is measured on.
Test models before release, tag their errors (spelling, legal term, name, number, code-mixing, OCR, wrong generated answers)
When users report issues with transcription quality, triage them to identify the most likely cause and recommend the fix.
5. Annotations
Write annotation guidelines and train and guide our part-time annotators.
Qualifications
Must have
A postgraduate degree in Linguistics, Hindi, or a related language field.
Hindi proficiency: native or near-native command of spoken and written Hindi, and write correct standard Hindi with accurate spelling, matras and punctuation.
Devanagari literacy and typing: fluent reading of Devanagari text, including legal documents, and the ability to type it accurately and at speed using any input method.
2+ years of transcription, subtitling, annotation or linguistic QC.
A sharp ear and real consistency: you apply a rule the same way on the 500th file as on the first.
Clear written English. You will write the guidelines other people follow.
Comfort conversing with people in Hindi and English and turning conversations into structured notes.
A fast learner who picks up new tools, guidelines and workflows quickly.
Open to managing people: coordinating annotators, reviewing their work and helping them improve.
Strong plus
Familiarity with regional Hindi varieties (Malvi, Bundeli, Bagheli, Rajasthani, Awadhi, Bhojpuri) or dialect fieldwork. Experience with Madhya Pradesh varieties is especially welcome.
Exposure to legal or court Hindi, including Urdu-origin legal vocabulary.
Working knowledge of Hindi phonology and regional phonological variation, and experience using it to define transcription or spelling conventions for Hindi datasets.
Experience reviewing or supervising other annotators’ work.
Experience with annotation tools such as Label Studio, ELAN, Labelbox, or Prodigy.
What You Will Achieve in a Year
Set up and run your own Hindi annotation workflows: how data comes in, who annotates it, how it is reviewed.
Write the Hindi style guide that our lawyers, annotators and ML team all work from.
Build the Hindi evaluation sets that every model release is judged against.
Map the Hindi varieties across our states, and close the gaps in our data.
Trace user-reported Hindi issues to their cause and turn them into fixes.
Introduction call: the role, your background, what you are looking for, and culture fit.
Take-home assignment to evaluate your skills.
Deep-dive with our Data or ML team.
Offer.
Benefits and Perks
WFH with flexible work hours.
Unlimited PTO.
Contacts within the Harvard / MIT/ Oxford ecosystem.
Autonomy and Ownership
Smart, Humble and Friendly peers
Generous vacation
Maternity and Paternity leaves
Learning & Development resources