Key Responsibilities

- Lead end-to-end SLM lifecycle (data ingestion - dataset creation - fine-tuning - evaluation - deployment)

- Architect scalable ETL pipelines for large-scale unstructured data

- Design and build training datasets (instruction tuning, classification, entity extraction)

- Implement fine-tuning techniques (LoRA, QLoRA, PEFT)

- Define evaluation frameworks (accuracy, F1, domain-specific KPIs)

- Optimize models using quantization (INT4/INT8), inference tuning

- Drive system design for SLM-powered solutions (APIs, pipelines, infra)

- Mentor engineers and guide best practices

Must-Have Skills :

- Strong Python (data processing, ML pipelines)

- Hands-on experience in unstructured text processing (NLP)

- Experience designing ETL pipelines (batch/streaming)

- Proven experience in SLM/LLM fine-tuning

- Understanding of GPU memory management (VRAM), model loading, batching, and throughput optimization

- Experience with CUDA-enabled environments / GPU clusters (NVIDIA GPUs such as A100, L40S, T4, etc.)

- Familiarity with deployment on GPU-enabled platforms (Kubernetes / cloud GPU instances / on-prem setups)

- Experience with dataset creation (JSONL, instruction tuning)

- Familiarity with Hugging Face, PEFT, Unsloth or similar

- Experience with vector DBs (pgvector / milvus etc)

- Experience building APIs (FastAPI/Flask)

Good to Have :

- Experience in RAG pipelines

- Knowledge of model optimization (quantization, pruning)

- Exposure to Ollama / on-device inference / GGUF models

- Experience with evaluation frameworks (RAGAS / OpenAI evals)

- Domain experience (telecom, logs, enterprise ops)

- Exposure to cloud platforms (Azure / AWS / GCP)