LLM Research Engineer - Pune, India

Atlassian; Pune District; On-site; INR 1,500,000 - 2,700,000; Full time

Job summary

Predii Inc. in Pune, India is seeking an LLM Research Engineer to own end-to-end model development, fine-tuning and deployment in production environments.

You will ship backend services, APIs, and integrated pipelines while collaborating with cross-functional teams and customers. You should bring 2-3+ years in production ML/NLP, strong CS fundamentals, and hands-on experience with Python or Java, Pandas, and modern ML frameworks.

Qualifications

• Bachelor's/Master's in CS or PhD in NLP/ML/LLM (thesis submitted)
• 2-3+ years in a production environment
• Rock-solid CS fundamentals — algorithms, data structures, OS
• Python or Java, you pick
• Comfortable with Pandas and data analysis

Responsibilities

• Own end-to-end model development, from design to deployment
• Build and fine-tune models for Q&A and real-world use
• Ship backend services (microservices, REST APIs) and data pipelines
• Collaborate with cross-functional teams, customers, and open-source community

Skills

Python

Java

Pandas

Data analysis

CS fundamentals

NLP/LLM fundamentals

Education

Bachelor's/Master's in CS

PhD in NLP/ML/LLM

Tools

PyTorch

Keras

scikit-learn

spaCy

Elasticsearch

MongoDB

HDFS

Git

Jira

Confluence

Gradle

Maven

Job description

Predii is hiring an LLM Research Engineer

Pune, India | Competitive Salary | 2‑3 Years Experience

Who we are

Predii's an Enterprise AI company (Palo Alto + Pune) that's been quietly crushing it in a niche nobody talks about — turning messy, unstructured repair data (2B+ jobs a month, no cap) into predictive insights for the automotive aftermarket. Big names like Snap-on, Epicor, Mercedes-Benz, and Valvoline trust our AI engine. Gartner and ABI Research have given us their stamp of approval too. We've been doing NLP for 8+ years — before it was cool.

The role, no fluff

You'll be deep in the LLM trenches — building and fine‑tuning models for Q&A, shipping backend services (microservices, REST APIs, the whole stack), and making sure our models actually work in the real world, not just in a notebook. You're not just doing research for the sake of research — you're the person who owns it end‑to‑end: design it, build it, ship it, improve it. You'll work with cross‑functional teams, customers, and the open‑source community, and get to help build genuinely industry‑changing stuff. Individual contributor energy, but you're never actually alone.

What you need to bring

• Bachelor's/Master's in CS (or a PhD in NLP/ML/LLM — thesis submitted? still counts)
• 2‑3+ years in a production environment (PhDs with a strong pub record at a top lab, you're exempt)
• Rock-solid CS fundamentals — algorithms, data structures, OS, the classics
• Python or Java, you pick
• Comfortable with Pandas and data analysis

Nice‑to‑haves (bonus points, not dealbreakers)

• You've messed with multimodal systems
• You know your way around embeddings — TF‑IDF, Word2Vec, GloVe, BERT
• Deep learning fundamentals: RNNs, LSTMs, CNNs, Attention, seq2seq — you get it, not just buzzwords
• NLP basics on lock: n‑grams, NER, POS tagging, semantics/syntax
• You've wrangled unstructured data and clustering algorithms before
• spaCy? You've got range
• NoSQL experience — Elasticsearch, MongoDB, HDFS

Extra flex if you have it

• LLM fine‑tuning, prompt engineering, eval frameworks
• Transformers, embedding models, Hugging Face publishing/inference
• PyTorch, Keras, scikit‑learn
• ML breadth — forecasting, classification, decision trees, random forests, neural nets
• Stats chops — regression, time series, topic modeling
• You know your way around Git, Jira, Confluence, Gradle/Maven