Careers

Applied ML Engineer, Small Models

Fine-tune domain models on BFSI data with QLoRA, run the evaluation harness, and get inference running inside customer perimeters.

Objective

Join our Model Layer team to fine-tune compact, domain-specialized models for BFSI use cases. You'll own the full ML lifecycle — data curation, QLoRA fine-tuning, evaluation, and deploying inference inside customer perimeters (VPC / on-prem).

About the role

Most BFSI work is narrow and well-specified — exactly what a small, fine-tuned model is good at. Your job is to make 1B–13B models carry real regulatory work: field extraction, classification, routing, and domain validation, running inside the customer's own perimeter with no cloud to fall back on.

You own the whole loop — data, training, evaluation, serving — and you document every checkpoint so a model shipped in March can be defended in September. The registry is not a suggestion; it is the product.

Key responsibilities

Fine-tune small language models (1B–13B) using LoRA / QLoRA / PEFT techniques
Curate, clean, and version high-quality BFSI training datasets
Build and maintain rigorous evaluation harnesses for domain accuracy
Optimize models for inference (quantization, distillation, pruning)
Deploy and support models running inside customer environments (air-gapped / VPC)
Partner with Solutions Architects to align model behavior with customer requirements

Required skills

2–5 years of experience in Applied ML / NLP
Strong Python, PyTorch, and Hugging Face (Transformers, PEFT, TRL, Datasets)
Hands-on with QLoRA, LoRA, SFT, and DPO/RLHF
Experience with model quantization (bitsandbytes, GPTQ, AWQ, GGUF)
Familiarity with inference servers (vLLM, TGI, Ollama, Triton)
Solid understanding of evaluation methodology (benchmarks, LLM-as-judge, human eval)
Experience with GPU workflows (CUDA, distributed training)

Nice to have

BFSI domain knowledge (banking, insurance, capital markets)
Experience deploying models in regulated / on-prem environments
Published work, Kaggle rank, or OSS contributions in NLP/LLMs

Skills & technologies

PythonPyTorchHugging FaceQLoRA / LoRA / PEFTSFT / DPOQuantizationvLLM / TGI / TritonEvaluation methodologyCUDA

Apply for this role Back to all roles

Rather talk than apply?

Tell us what you would build and why it matters. If the timing is right, the conversation goes from there.

Send a note

About the role

Key responsibilities

Fine-tune small language models (1B–13B) using LoRA / QLoRA / PEFT techniques

Curate, clean, and version high-quality BFSI training datasets

Build and maintain rigorous evaluation harnesses for domain accuracy

Optimize models for inference (quantization, distillation, pruning)

Deploy and support models running inside customer environments (air-gapped / VPC)

Partner with Solutions Architects to align model behavior with customer requirements

Required skills

2–5 years of experience in Applied ML / NLP

Strong Python, PyTorch, and Hugging Face (Transformers, PEFT, TRL, Datasets)

Hands-on with QLoRA, LoRA, SFT, and DPO/RLHF

Experience with model quantization (bitsandbytes, GPTQ, AWQ, GGUF)

Familiarity with inference servers (vLLM, TGI, Ollama, Triton)

Solid understanding of evaluation methodology (benchmarks, LLM-as-judge, human eval)

Experience with GPU workflows (CUDA, distributed training)

Applied ML Engineer, Small Models

Objective

About the role

Key responsibilities

Required skills

Nice to have

Skills & technologies

Role at a glance

Rather talk than apply?

Applied ML Engineer, Small Models

Objective

About the role

Key responsibilities

Required skills

Nice to have

Skills & technologies

Role at a glance

Rather talk than apply?