Remote-first (India)
India has more than 50 million pending court cases. Behind that number are people waiting years for decisions that affect their lives. Courts spend enormous time on manual and clerical work: recording proceedings, preparing orders, reading lengthy files, moving documents, and keeping fragmented systems in sync. Judges should be spending more of that time delivering justice.
Our mission at Adalat AI is to make timely justice possible. We are a Y Combinator–backed nonprofit building an end-to-end technology stack for justice systems. We want to do for courts what UPI did for payments: create a common, reliable technology layer on which an entire ecosystem can work. That means going beyond individual AI features to build the infrastructure through which courts can operate more efficiently, securely, and accessibly.
Our products span AI transcription in Indian languages, dashboards for judges and court administrators, WhatsApp assistants for litigants and advocates, AI-powered workflows for courts and registries, and a Paperless Courts platform for state judiciaries. Adalat AI’s technology is used in more than 5,300 courtrooms across 10 Indian states—roughly 25% of India’s judiciary. Our next ambition is to reach every courtroom in the country, and eventually help justice systems across the Global South.
Our work has been featured in a Stanford case study and The Wall Street Journal, and recognised by the Harvard President’s Innovation Challenge, MIT Solve, and the AI Innovation Challenge at the Global AI Summit. Incubated at MIT and Oxford, we are supported by leading technology companies and funders.
Founded by a team with backgrounds in law, technology, and economics from Harvard, Oxford, MIT, and IIIT Hyderabad, we are now building the teams and technology needed for that next stage of scale.
Role Overview
Why this role matters
A court’s work is built around language. A judge may need to understand a 700-page case file before a hearing. An advocate may need to find the right passage in dozens of judgments. A registry may have to check hundreds of filings against detailed procedural requirements. Much of the material is scanned, handwritten, multilingual, poorly structured, and scattered across documents that were never designed for machines to read.
General-purpose language models are powerful, but legal work demands much more than plausible answers. A system must retrieve the right evidence, distinguish facts from arguments, follow references across documents, preserve the meaning of legal terms across languages, and show exactly where an answer came from. Missing one qualification, citing the wrong paragraph, or inventing a fact can make an otherwise fluent response useless.
We have already started building this layer. Legal Lens helps users ask questions of case documents and receive citation-grounded answers, combining document parsing, retrieval, and an agentic question-answering system. Our text work also includes legal translation, writing aids, summarisation, and document understanding. On the horizon are harder problems: multi-document and multimodal reasoning, handwritten and degraded filings, structured case intelligence, scrutiny workflows in Paperless Courts, and domain-specific copilots such as Customs Copilot.
Taking these systems from individual features to dependable infrastructure for courts across India requires a step change in language-model capability, evaluation, data, and team strength. We are hiring a Head of Language Technology as a foundational leader in that next phase.
The role
You will build and lead Adalat AI’s language technology team: the researchers and ML engineers working on NLP, large language models, retrieval, legal reasoning, multilingual text, and document intelligence. You will set the technical direction, recruit and mentor exceptional people, and own the path from research results to language capabilities that can be trusted in production.
Your scope is the language and document intelligence side of AI, complementary to our speech technology team. You will directly lead language researchers and ML engineers, working closely with—but not directly managing—the runtime, platform, and general software engineering teams that help put these systems into production.
This is a role for someone who has led serious language-model or NLP work at an established AI organisation, research lab, or product company—and shipped systems used at meaningful scale. You should be comfortable reviewing a paper, designing a post-training experiment, challenging an agent architecture, diagnosing a failure in a real case file, and deciding what to build next.
This is a fast-moving, high-ownership environment. We operate where clean benchmarks, ready-made datasets, and complete specifications often do not exist. You will need to make sound decisions with incomplete information, stay close to real users, and move quickly without compromising correctness, security, or institutional trust.
You will work with our founders, the Head of AI & ML, speech and runtime leaders, product and engineering teams, legal experts, data operations, and colleagues working directly with courts. You will help define both the technology and the research organisation needed to scale it.
What success looks like
Within your first six months, you will have established a clear language technology roadmap and built an effective team with explicit ownership. You will have shipped measurable improvements in at least one major legal-text workflow, strengthened our evaluations for answer quality and citation reliability, and established a practical approach to data generation and post-training. You will have identified the core capabilities and bottlenecks that determine how language AI scales across Adalat AI’s products.
Over time, success means an organisation that can continuously improve legal question answering, document intelligence, translation, summarisation, and reasoning across languages and workflows. Courts should be able to find relevant information faster, spend less time preparing and checking documents, and rely on AI assistance that is grounded, explainable, and safe enough for consequential work.
We will judge success by the quality of the team you build and whether the systems make real work in courts easier—not by the number of demos, papers, or model releases alone.
Key Responsibilities
Language technology strategy. Set a multi-year research and product roadmap across legal NLP, LLMs, retrieval, multilingual generation, legal reasoning, and document intelligence. Decide where frontier models are sufficient, where domain adaptation matters, and where original research is necessary.
An exceptional language ML team. Recruit, mentor, and develop language researchers and ML engineers. Grow technical leads, establish rigorous research practices, and build a team capable of taking an ambiguous field problem through experimentation, evaluation, and a production release. This is an early leadership hire in our wider ML and engineering expansion.
Legal Lens and grounded question answering. Improve systems that answer questions over complex legal documents with precise, verifiable citations. Tackle long context, retrieval quality, multi-hop reasoning, completeness, contradictory evidence, document navigation, and agent reliability. Extend from one document to collections of filings, statutes, orders, and judgments.
Legal language models and post-training. Lead domain adaptation, supervised fine-tuning, preference optimisation, reinforcement learning where useful, synthetic data generation, distillation, and model selection. Establish a repeatable data-to-post-training loop that improves real legal tasks—not just general benchmark scores.
Multilingual language understanding. Build high-quality translation, summarisation, rewriting, and information extraction for Indian languages and legal English. Handle code-mixed text, legal terminology, low-resource languages, and situations where subtle differences in wording change the meaning of a document.
Document intelligence and multimodal understanding. Advance OCR and document parsing for scanned, handwritten, degraded, and layout-heavy case files. Develop systems for tables, annexures, forms, exhibits, structured information extraction, and reliable reasoning over documents that combine text and visual information.
New legal AI workflows. Partner with product teams on document scrutiny, legal research, case timelines, precedent retrieval, structured case histories, and copilots for judicial and government workflows—including the research behind Customs Copilot. Distinguish capabilities ready to deploy from promising but unproven research.
Evaluations that reflect legal consequences. Develop representative benchmarks with lawyers and language experts. Measure correctness, completeness, citation faithfulness, grounding, translation fidelity, hallucinations, and real user correction effort. Build failure analysis and feedback loops that turn mistakes in the field into better data and models.
Research that survives production. Partner with runtime, platform, and product engineers on latency, inference cost, privacy, on-premises deployment, observability, release quality, and reliability. Own language-model and system-quality decisions while collaborating across engineering boundaries.
Research contributions and technical leadership. Set the bar for reproducible experimentation and thoughtful publication or open-source work where appropriate. Build external research relationships and help make Adalat AI a place where strong researchers want to do their most important work.
Qualifications
At least 10 years of relevant experience across NLP, machine learning, or language-model research and engineering, with substantial technical depth and a record of shipping real products.
At least 3 years of direct people management, including hiring, coaching, evaluating, and developing strong NLP or language-model researchers and engineers through multiple releases.
Deep technical judgment in modern language models. You understand transformers, pretraining and post-training choices, fine-tuning, retrieval, agent architectures, and evaluation. You can reason from first principles about why a model or system fails.
Evidence of production impact. You have helped build or lead language-based products used at meaningful scale and can explain the trade-offs behind model quality, latency, reliability, and cost.
Experience with genuinely hard data problems. You have worked where the data distribution is messy or difficult to reproduce, high-quality labels are scarce, and standard benchmarks miss the failures that matter.
Hands-on research fluency. Strong grounding in Python and modern ML tooling such as PyTorch and Hugging Face, with the ability to inspect code, review experiments, and guide architectural decisions.
Research credibility. A strong record of original technical contributions, ideally including publications at ACL, EMNLP, NAACL, NeurIPS, ICML, ICLR, or comparable venues. You know how to turn research insight into something people can use.
Education. A B.Tech or equivalent undergraduate degree and a master’s degree in computer science, machine learning, computational linguistics, or a related field. A PhD is strongly preferred.
Leadership across disciplines. You can lead researchers and ML engineers while collaborating with separate product, runtime, platform, legal, and data teams. You can maintain a high bar without becoming the bottleneck.
Comfort with pace and ambiguity. You thrive in a fast-moving environment where teams learn directly from users and must continually adapt their approach.
A reason to care about the problem. You want your technical work to contribute to public infrastructure and are willing to learn from judges, lawyers, court staff, and other people closest to the problem.
Experience in multilingual or low-resource NLP, legal or other high-stakes domains, long-document reasoning, OCR/vision-language models, retrieval systems, model efficiency, or privacy-sensitive deployments would be especially valuable. Significant open-source contributions or leadership of widely used language AI products are compelling evidence of depth.
Know more about Adalat AI
Why Adalat AI
Some problems are technically hard. Others have the potential to change millions of lives. It is rare to work on one that is both—and rarer still to do it alongside people who care as deeply about the mission as you do.
At Adalat AI, we are trying to build the infrastructure that makes justice work at scale. We believe courts can have their UPI moment: a shared technology layer that transforms how an entire system operates. If we succeed, the work we do today in Indian courts could become part of public infrastructure serving billions of people over time.
Language technology is central to that possibility. The hardest problems are not simply making an LLM answer a question; they are helping it understand real case files, find and cite the right evidence, reason over complex histories, and work across languages and document formats that mainstream AI often neglects. You will be building capabilities that people can use in the institutions they depend on—not just another chatbot.
You will work alongside ambitious researchers, engineers, designers, lawyers, and people close to courts who share the same mission. The team values technical depth, curiosity, thoughtful disagreement, and the willingness to get work into users’ hands. There is room to publish and push research forward, but also a direct path to seeing what your work changes in the world.
If you have ever wanted to use your technical skills at the intersection of hard problems, exceptional teammates, and meaningful impact, Adalat AI is a place to do it.
To start a conversation, share your CV or profile and a short account of a language AI system you led: the technical challenge, the team, what reached production, and what improved.