About the role

We’re building AI and automation products on open models, and we need someone who can make them trustworthy. You’ll work at the model level to understand why models hallucinate and engineer systems that keep outputs grounded, accurate, and controllable.

This is a hands-on role for someone who would rather fix the underlying problem than patch it with a prompt.

What you’ll do

- Fine-tune, evaluate, and deploy open-weight models (Llama, Mistral, Qwen, and similar) for production use
- Research and implement methods to reduce hallucinations: retrieval grounding, constrained decoding, calibration, preference optimization, and verification layers
- Build evaluation pipelines and benchmarks that measure factuality and reliability, not just fluency
- Run red-teaming and failure analysis to find where models break, and close the gaps
- Work with product and engineering to ship reliable model behavior in real applications
- Stay current with recent open model releases and techniques, and tell us which are worth adopting

What we’re looking for





- Strong experience with modern ML and LLMs: PyTorch, Hugging Face, and fine-tuning methods such as LoRA/QLoRA, SFT, and RLHF/DPO
- A track record of improving model reliability, whether through RAG, uncertainty estimation, guardrails, or evaluation design
- Comfort running inference and training on open models, including quantization and serving
- Solid engineering habits: clean code, reproducible experiments, clear documentation
- The ability to explain technical trade-offs to non-technical teammates

Bonus: robotics experience Knowledge of robotics is a strong plus, especially in perception, embodied AI, vision-language-action models, simulation, or ROS. We’re interested in how reliable model behavior carries over from text to systems that act in the physical world.

Why join us

- Work on hard, unsolved reliability problems with open models
- High ownership and real influence in a small, fast-moving team

📌 Machine Learning Specialist (India)
🏢 Ju0026W Nexus
📍 India