Harshwal Consulting Services LLP
Website:
harshwalconsulting.com
Job details:
Experience
2 - 3 years
About The Role
We're looking for a skilled LLM Engineer with a solid data science foundation to design, build, and maintain systems leveraging large language models — turning cutting-edge capabilities into reliable, scalable product features.
Key Responsibilities
- Design and implement LLM pipelines: prompt engineering, RAG, and fine-tuning workflows.
- Build, train, and evaluate ML/DL models for classification, regression, and clustering tasks.
- Develop NLP pipelines: NER, text classification, summarization, and sentiment analysis.
- Perform EDA, feature engineering, and statistical modelling on structured/unstructured data.
- Integrate LLM APIs (OpenAI, Anthropic, Mistral, open-source) into production services.
- Collaborate with backend engineers to serve models at scale with appropriate guardrails.
- Build tooling for model evaluation, A/B testing, and iterative prompt improvement.
Foundational Skills — Ml, Dl & Nlp
Machine Learning
Scikit-learn XGBoost / LightGBM Pandas / NumPy Hyperparameter tuning
- Core algorithms: regression, decision trees, random forests, SVMs, and ensembles.
- Full ML lifecycle: data cleaning, feature engineering, training, evaluation, and deployment.
- Evaluation metrics: F1, AUC-ROC, RMSE based on task type. Cross-validation best practices.
Deep Learning
PyTorch TensorFlow / Keras Transformers LoRA / PEFT GPU training
- Build and train neural networks — CNNs, RNNs, LSTMs, and Transformer architectures.
- Transfer learning and fine-tuning with LoRA/PEFT. Mixed-precision GPU training.
- Attention mechanisms, positional encoding, and multi-head attention fundamentals.
Natural Language Processing
Hugging Face spaCy / NLTK BERT / GPT Semantic search Embeddings
- NLP fundamentals: tokenisation, stemming, POS tagging, NER, dependency parsing.
- Word2Vec, GloVe, FastText, and contextual embeddings (BERT, sentence-transformers).
- Text classification, summarisation, Q&A, and sentiment pipelines in production.
- Semantic search, dense retrieval, and embedding-based similarity for RAG systems.
Llm-specific Skills
- 2-3 yrs experience, with 1+ year hands-on with LLMs.
- Prompt engineering, few-shot learning, chain-of-thought, and instruction tuning.
- RAG pipelines with vector DBs (Pinecone, Weaviate, Chroma, pgvector).
- LLM orchestration: LangChain or LlamaIndex.
- Open-source models via Ollama / vLLM for local inference.
- REST APIs and scalable Python backend services.
- Cloud platforms: AWS, GCP, or Azure.
Click on Apply to know more.