Sr. LLM Fine Tuning Engineer

BCI has an open remote position on our GenAI team working with our USA based client. The LLM Fine Tuning Engineer will join our offshore development team that is growing and there is a lot of new and exciting GenAI work to be completed. This is a full-time remote position and must be able to work blended hours of EST / IST timings. Position will have a mix of LLM Fine tuning and hands-on Python and GenAI development.

About the Role

We're looking for an LLM Fine Tuning Engineer with python and GenAI background to work on customization and optimization of large language models for production use cases. This role owns the full fine-tuning lifecycle — from data preparation through training, evaluation, and deployment — working with open-weight models (e.g., Llama, Gemma) as well as proprietary/managed models (e.g., Google Gemini) where fine-tuning access is available. You'll be hands-on with real training runs at scale, not just prompt engineering, or API integration.

What You'll Do

• Fine-tune and adapt large language models (Llama, Gemma, and other open-weight models, plus managed options like Gemini where applicable) for specific business use cases
• Design and execute full fine-tuning pipelines: dataset curation and cleaning, tokenization, training/eval splits, hyperparameter selection, and training runs (full fine-tune, LoRA/QLoRA, PEFT, RLHF/DPO as appropriate)
• Run and manage large-scale training jobs across multi-GPU / distributed environments
• Evaluate model performance using both automated benchmarks and human-in-the-loop review; iterate to close quality gaps
• Optimize models for production inference (quantization, distillation, latency/cost tradeoffs)
• Deploy fine-tuned models into production systems and monitor performance, drift, and degradation over time
• Collaborate with data, ML infrastructure, and product teams to define fine-tuning objectives and success metrics
• Stay current on the open-model landscape and evaluate new base models as candidates for fine-tuning
• Document methodology, training runs, and results for reproducibility and knowledge sharing

Required Qualifications

• Strong proficiency in Python, with solid software engineering fundamentals (not just notebooks)
• Hands-on, production-level experience fine-tuning LLMs — this is a must-have, not exploratory/academic experience only
• Demonstrated experience taking fine-tuned models into live, large-scale production systems (not just POCs)
• AWS and /or Google cloud production experience will be considered
• Experience with open-weight model families (e.g., Llama, Gemma, Mistral, or similar)
• Practical knowledge of fine-tuning techniques: LoRA/QLoRA, PEFT, full fine-tuning, instruction tuning, RLHF/DPO
• Experience with ML/training frameworks such as PyTorch, Hugging Face Transformers/TRL/PEFT, DeepSpeed, or similar
• Familiarity with distributed/multi-GPU training and the associated infrastructure challenges
• Solid understanding of model evaluation methodology for generative models

Nice to Have

• Experience fine-tuning or customizing Google Gemini or other managed/API-based models
• Experience with vector databases, RAG architectures, or hybrid RAG + fine-tuning approaches
• Experience with MLOps tooling for training pipelines (e.g., MLflow, Weights & Biases, Kubeflow, SageMaker, Vertex AI)
• Experience with model quantization and inference optimization (vLLM, TensorRT-LLM, GGUF, etc.)
• Background in NLP research or publications related to LLM training/fine-tuning

What Success Looks Like

Within your first few months, you're independently running fine-tuning jobs on open models, have a clear point of view on which base models and techniques fit which use cases, and have shipped at least one fine-tuned model into a production system with measurable quality improvement over baseline.

Interview Process:

1. If profile appears to fit role, we will send you a request for more information and details on your background. 2. Initial 30 min MS Teams conversation with BCI-IT team to go over your hands on experience and determine fit. 3. If potential fit, you will be sent a video technical screen with 8 questions on Python, GenAI and LLM fine tuning. 4. 45 min to 1 hour client technical interview with Python code share activity. You will be speaking with 2-3 Sr. team members. Hiring decision can be made after call.