Machine Learning Engineer – AI Research

Lexapar Analytics Private Limited; Bengaluru; On-site; INR 1,200,000 - 2,400,000; Full time

Benefits offered by this job

Competitive compensation

Collaborative environment

Job summary

Lexapar Analytics Private Limited is seeking a Machine Learning Engineer to build the AI core of the Lexapar platform in Bengaluru. This high‑ownership role demands hands‑on model‑building, rigorous evaluation, and productive collaboration with product and legal domain experts to translate complex workflows into production‑grade legal AI solutions.

You will train and fine‑tune deep learning models, design robust evaluation benchmarks, and implement RL and RLHF pipelines while balancing model

Qualifications

• Hands-on DL experience in production or research settings.
• Fine-tuning large language or DL models with data curation.
• Experience designing model evals, benchmarks, metrics & error analysis.
• Knowledge of RL techniques (RLHF, PPO, DPO) and practical use.
• Strong software engineering fundamentals and problem solving.
• Experience tracking experiments with Weights & Biases or MLflow.

Responsibilities

• Model Training & Fine‑Tuning: Train and fine‑tune DL models on domain‑specific legal data.
• Evaluation & Benchmarking: Design and run evals to measure model quality and legal accuracy.
• Reinforcement Learning: Build RL/RLHF pipelines to align model behavior with legal correctness.
• Research & Experiment Tracking: Run experiments and maintain reproducible results.
• Design & Problem‑Solving: Scope ambiguous problems and balance quality, latency, and cost.
• Cross‑Functional Collaboration: Work with product, engineering, and legal experts to meet standards.

Skills

Deep learning

Fine-tuning

Reinforcement Learning

Model evaluation

Python

Experiment tracking

Tools

Weights & Biases

MLflow

Job description

Build the AI core of the Lexapar platform

We are seeking a Machine Learning Engineer with a strong foundation in deep learning and hands‑on model-building experience to help build the AI core of the Lexapar platform. This is a high‑ownership role for someone who is equally comfortable running rigorous evaluations, training and fine‑tuning models, and pushing the frontier with reinforcement learning and who enjoys the full arc from open‑ended problem design to production‑grade results. You will work closely with product and legal domain experts to turn hard, ambiguous problems in legal AI into models that meet the accuracy and reliability bar the profession demands.

What you’ll do

Model Training & Fine‑Tuning: Train and fine‑tune deep learning and large language models on domain‑specific legal data, iterating quickly across architectures, datasets, and training regimes.

Evaluation & Benchmarking: Design and run rigorous evals to measure model quality, safety, and legal accuracy; build the harnesses, datasets, and metrics needed to track progress objectively.

Reinforcement Learning: Build and train RL and RLHF/preference‑optimization pipelines to align model behavior with legal‑domain correctness and user intent.

Research & Experiment Tracking: Run structured experiments, track results systematically, and maintain a clear, reproducible record of what was tried, what worked, and what didn’t.

Design & Problem‑Solving: Independently scope ambiguous research and engineering problems, propose approaches, and make sound trade‑off calls between model quality, latency, and cost.

Cross‑Functional Collaboration: Partner with product, engineering, and legal domain experts to translate real workflow needs into model requirements, and to validate outputs against professional legal standards.

Qualifications

Deep Learning Experience: 3+ years of hands‑on experience building, training, and deploying deep learning models in production or research settings.

Training & Fine‑Tuning: Demonstrated experience fine‑tuning large language or other deep learning models, including data curation, training infrastructure, and iteration at scale.

Evaluation Rigor: Practical experience designing and running model evals including benchmark construction, metric design, and error analysis.

Reinforcement Learning: Working knowledge of RL techniques (e.g., RLHF, PPO, DPO, or similar) and experience applying them to real models.

Engineering & Problem‑Solving: Strong software engineering fundamentals with the ability to independently design solutions to open‑ended, technically ambiguous problems.

Research Discipline: Comfortable tracking experiments and results systematically (e.g., via tools like Weights & Biases or MLflow) and communicating findings clearly.

Autonomy & Drive: Comfortable operating with a high degree of independence, initiative, and speed in a fast‑moving, resource‑lean environment.

Domain Context (Preferred): Interest in or exposure to legal technology, NLP for regulated domains, or similarly high‑stakes application areas.

What we offer

The opportunity to build the foundational AI models behind a category‑defining legal tech product, with direct visibility into company strategy.

A collaborative environment working alongside experienced technologists and legal professionals who speak your language.

Competitive compensation, commensurate with experience.

About Lexapar

At Lexapar, we're reimagining how legal work gets done - combining powerful Al with human expertise to make contracts faster, smarter, and more accessible. We're building technology that doesn't just automate processes, but truly transforms how businesses, lawyers, and professionals think about legal delivery.

If you're excited by the idea of shaping the next generation of legal technology — whether you're a lawyer, engineer, data scientist, or business strategist — Lexapar is the place where your ideas can have real impact. We value curiosity, clarity, and courage — and we're always looking for driven individuals who want to build something meaningful.

We're growing fast and always open to connecting with exceptional people who share our vision. If you'd like to explore opportunities to work with us.