CTS India

Mumbai, Maharashtra, India

AI Data Engineer LLM RAG (Junior)

Job ID JOB-005-0000238

Team: BIM-AI Core Reports to: SEM, BIM-AI Core

The models need to be deployed, the RAG index needs to work, and every project that runs through the platform should make the next one cheaper. This role supports the models, the data flywheel, and the fine-tuning that improves BoQ tolerance over time, working under the guidance of senior ML engineers on the BIM-AI Core team.

What you will do

• Support deployment and operation of DeepSeek R1 32B and Mistral Large 3 on the sovereign stack, under guidance.
• Help build the RAG pipeline over CTS DC project history — BoQ data, D&B rates, specifications.
• Support evaluation harnesses with the QA Lead: BoQ tolerance, layout quality, retrieval precision and recall.
• Support supervised fine-tuning on CTS project data, working towards the ±8% tolerance target.
• Help ensure data sovereignty holds in practice, not just on paper.

Must have

• 2–5 years in applied ML or data engineering, with some exposure to deploying LLMs.
• Some hands-on experience with open-weight models (DeepSeek, Mistral, Llama) and inference stacks (vLLM, Ollama, TGI, llama.cpp).
• Working knowledge of RAG fundamentals: chunking strategies, embeddings, vector DBs, retrieval evaluation, re-ranking.
• Exposure to operating on-prem GPU hardware.

Nice to have

• Supervised fine-tuning (LoRA, QLoRA, or full fine-tune) on 30B+ models.
• AEC or construction background.
• DPDP Act or EU AI Act compliance experience.

Skills LLM deployment · RAG · Fine-tuning (LoRA/QLoRA) · vLLM / Ollama · Vector databases · Python · On-prem GPU

Role Details Models: DeepSeek R1 32B · Mistral Large 3 Stack: vLLM · Qdrant · Python Target: BoQ tolerance ±8% via supervised fine-tuning

Required Skills

Desired Skills