Urgent Opening For Senior Machine Learning Engineer (NLP)- Pune

Dear Candidate,

Please find the job description below,

Senior Machine Learning Engineer (NLP)

An experienced, production-focused Machine Learning Engineer specializing in Natural Language Processing (NLP) and high-performance backend systems for the academic publishing domain.

Responsible for the full ML lifecycle — from data labeling and model training to deploying concurrent, production-grade inference APIs that process complex textual data at scale.

1. Advanced Natural Language Processing & Information Extraction

Building, extending, and training custom spaCy (v3.6+) pipelines and components (NER, SpanCat, SpanFinder, custom tokenizers). Experience with SOTA frameworks such as Thinc, Flair, and zero-shot architectures like GLiNER. Knowledge of embedding generation, vector spaces, and modern embedding models (e.g., google/embeddinggemma-300m). Mastery of fuzzy matching (Levenshtein distance), regex, and structured data formats (XML, PDF, DOCX, LaTeX), with a focus on academic manuscript structures and metadata standards.

2. Deep Learning & Model Optimization

Proficiency in PyTorch 2.x and PyTorch Lightning for reproducible model training across hardware accelerations (CUDA, MPS, CPU). Hands-on experience with the Hugging Face ecosystem for transfer learning and fine-tuning of LLMs and encoder-based transformers. Solid foundation in statistics, numerical computing (NumPy, SciPy, Pandas), and classical ML algorithms (e.g., DBSCAN clustering, Scikit-Learn pipelines).

3. High-Performance ML Operations & Backend Engineering

Advanced Python development using Asyncio and FastAPI to build high-throughput, low-latency REST APIs. Understanding of the GIL, multi-threading, and memory/thread safety when serving heavy ML models in production web servers.

4. MLOps, Cloud & Data Lifecycle

Managing the data lifecycle with tools like Prodigy and spaCy Projects for active learning and gold-standard datasets. Using MLflow for experiment tracking, model registry, and reproducibility. Deploying and managing applications on GCP (Compute Engine, GKE, Vertex AI, Gemini API). Writing robust test suites (pytest, pytest-asyncio) and managing CI/CD via GitHub Actions, Jenkins, Docker, and Kubernetes. Monitoring via ELK and Dynatrace, and handling data streams via Kafka.

Core

• Python 3.7+

• FastAPI, Uvicorn, Starlette, Pydantic v2 + pydantic-settings

• pytest, pytest-asyncio (unit + functional tests)

• asyncio, aiohttp

• Project packaging (setuptools + pyproject.toml)

• REST API development

ML

• PyTorch 2.x (CPU/GPU/MPS)

• CUDA 12.*+

• PyTorch Lightning

• Hugging Face Transformers

• Sentence_transformers

Classical ML

• scikit-learn, NumPy, Pandas, SciPy, matplotlib

• JupyterLab — exploration and training notebooks

• Basic knowledge of Linear Algebra

NLP

• Custom NLP pipeline design (spaCy, incl. transformer-encoder pipelines)

• Custom spaCy components: NER, SpanCat, Dependency parser, Sentencizer, SpanFinder, custom tokenizers/matchers

• Thinc, Flair NLP, GLiNER

• Retrieval and embedding models (e.g., google/embeddinggemma-300m)

Data processing

• spaCy Projects

• Prodigy

Text processing

• Fuzzy search

• Regex

• XML processing

LLM

• LLM-assisted data annotation with quality guards

• LLM inference servers: vLLM, Hugging Face TGI, llama.cpp

• Async Python (AsyncOpenAIClient) for high-throughput dataset processing

Cloud & integration

• Docker

• GitHub Actions

• GCP: Artifact Registry, GCS, Compute Engine, GKE, Logging, Vertex AI, Gemini API

• Jenkins, K8s, Dynatrace, ELK

• Apache Kafka

• MLflow

• Git

General knowledge

• NER and span classification

• Sequence labeling and document-level classification

• Transfer learning / fine-tuning transformers

• Train/eval/deploy lifecycle

• Data augmentation, information extraction, relation extraction, entity linking, information retrieval

• Fuzzy matching, clustering (DBSCAN)

• Precision/recall tradeoffs in information extraction

• Thread safety serving models in multi-threaded web servers

• XML processing; basic knowledge of doc(x), LaTeX, PDF formats

• Academic publishing domain — manuscript structure, metadata standards

Other tech as a plus

• Java, JavaScript