Staff Machine Learning Engineer

Role Overview

As a Staff Machine Learning Engineer, you will play a central role in driving the core ML research and engineering at Adalat AI. You will work across the ML lifecycle — from data design to training and deployment — and serve as a technical mentor to a growing team of ML engineers and researchers.

This role is ideal for someone with deep experience in training large models, especially in low-resource settings, and who thrives on ownership, autonomy, and real-world impact. You will help build systems that touch millions of lives by improving the functioning of the world's largest court system.

Key Responsibilities

Research & Systems Building

• Design, train, and deploy models for speech recognition, summarisation, legal Q\&A, retrieval, and translation.
• Build scalable ML systems using LLMs, transformers, and custom architectures.
• Train large models from scratch (or from base checkpoints) when needed, including curating and managing data pipelines.
• Contribute to original research; submit to top-tier conferences (A\*STAR/CORE-ranked such as ACL, NeurIPS, ICML, EMNLP, or similar).

Technical Leadership

• Mentor junior engineers and researchers on ML design, experimentation, and deployment practices.
• Lead technical design discussions and decisions on modeling strategies, data pipelines, and infrastructure.
• Set up best practices for reproducibility, evaluation, and documentation across ML projects.

Cross-functional Collaboration

• Translate product and legal requirements into technical architecture and model specs.
• Work with linguists, annotation teams, and legal domain experts to define data needs and ensure model reliability.
• Collaborate with backend engineers to ensure seamless integration of models into production systems.

About You

• Research Expertise: Strong background in AI research with a passion for applying advanced techniques to solve real-world problems. Experience handling the annotation team is a bonus.
• Leadership Ambition: Ready to step into a leadership role while maintaining hands-on involvement in research and development.
• Problem Solver: Ability to tackle complex technical challenges and develop innovative solutions.
• Collaborative Mindset: Excellent communication skills, humble attitude and ability to work cross-functionally with product and engineering teams.
• Startup Experience: Thrives in dynamic, fast-paced environments, preferably with experience in early-stage startups.
• LLM Expertise: Proven track record of building and shipping successful applications powered by Large Language Models.
• Customer-Centric Approach: Strong commitment to understanding and addressing customer needs through AI-driven solutions.

Qualifications

Ideal Profile

• PhD in ML, NLP, Speech, or a related field OR equivalent experience working on cutting-edge ML projects at scale.
• Experience publishing in top-tier A\*STAR-ranked AI/ML conferences (e.g., NeurIPS, ACL, EMNLP, ICML, CVPR, ICLR).
• Strong track record of building and deploying production-grade ML systems, ideally in low-resource or domain-specific environments.
• Proven experience training LLMs or ASR models from scratch, including building custom datasets and pipelines.
• Familiarity with ML system optimisation, including inference serving, model quantisation, and latency reduction.

Bonus: experience working in civic tech, public infrastructure, or legal-tech is highly appreciated.

You Might Thrive Here If You Are

• A hands-on builder and researcher, not afraid of messy data, ambiguous specs, or field deployments.
• A natural mentor, who enjoys helping others level up while maintaining high technical standards.
• Excited about justice tech and the chance to build systems that improve governance at population scale.
• Comfortable moving between experimentation and shipping, and between deep work and scrappy MVPs.

Nice to Have

• Experience with annotation team workflows and building training datasets in-house.
• Experience with retrieval-augmented generation (RAG), fine-tuning strategies, or few-shot learning.
• Familiarity with tools like Hugging Face Transformers, Weights & Biases, Ray, Triton, or ONNX.
• Background in legal, civic, or public policy work.