LLM Finetuning Expert at Lendenclub, Mumbai
Open. Posted .
Indic and financial language-model training spans continued pretraining, alignment, evaluation and low-latency inference for lending and credit products. The company describes itself as profitable and rapidly scaling.
Applying opens the employer's page. Want jobs like this sent to you each morning? Open Elip.
| Location | Mumbai, Maharashtra |
|---|---|
| Work mode | On-site |
| Experience | 3–4 years |
| Pay | Not stated by the employer |
| Role | AI Engineer |
| Skills asked for | Hugging Face Transformers, PEFT, TRL, RLHF, Reward modelling, DPO, GRPO, vLLM |
| Posted | 17 Nov 2026 |
Elip's summary
Adapts foundation language models for a financial technology company building credit and lending products. Trains and aligns Indic and code-mixed text models using Hugging Face Transformers, PEFT, and RLHF, then builds production inference with vLLM. Evaluates model robustness and numeric accuracy.
The job, in the employer's words
About VARTIS Platforms (Formerly known as LenDenClub Techserve Pvt. Ltd.)
At VARTIS Platforms, we are a corporate group on a mission to democratize credit access and empower millions of people through cutting-edge technology. We focus on building trust and convenience, leveraging AI to drive the information revolution for smarter risk management and operational efficiency. VARTIS Platforms is profitable, rapidly scaling, and industry-recognised. Our impact has been acknowledged with top honours, including being awarded as the Best Fintech Provider by Financial Express, a testament to our product excellence, customer experience, and leadership in regulated financial technology.
Our portfolio includes:
LenDenClub: An RBI-licensed P2P marketplace that connects borrowers with lenders in a seamless digital ecosystem.
InstaMoney: Instant digital loans that enable quick, secure, and accessible financial support.
NiveshGold: India’s next-gen digital vault for wealth preservation, starting with 100% pure, insured 24K digital gold.
UdharSaathi: A structured working capital platform built exclusively for QR active kirana stores, pharmacies, and small retailers.
Key Responsibilities
Select base models (Qwen, Llama, Gemma, Mistral, Indic models) and adapt tokenizers for Indic and code-mixed text.
Curate training data from conversations, emails and financial documents, including synthetic data from larger teacher models.
Run continued pretraining, SFT and LoRA/QLoRA or full fine-tuning for text and multimodal models.
Align models with RLHF, reward modelling, DPO or GRPO, including verifiable rewards for numeric finance tasks.
Train at scale on multi-GPU, multi-node setups using FSDP, DeepSpeed or Megatron with BF16/FP8.
Design evaluations covering accuracy, numeric hallucination, bias and robustness; find weaknesses and fix them.
Build scalable inference pipelines on vLLM, Triton Inference Server, SGLang or TensorRT-LLM, with quantisation for low-latency voice use.
Work with product, data and business teams to turn their problems into model specs and success metrics.
Mandatory Skills
3+ years in deep learning with models shipped to production, including at least one model of 1B+ parameters you trained end to end on multiple GPUs.
Hands-on with Hugging Face Transformers, PEFT and TRL, or equivalents such as Unsloth, Axolotl or torchtune.
Post-training beyond SFT (RLHF, reward modelling, DPO, GRPO or similar), with measured gains on held-out data.
Production inference experience on vLLM, Triton Inference Server or a similar engine.
Clear written and verbal communication: you can turn training results into a one-page recommendation a business leader can act on.
Preferred Skills
Debugging distributed training: divergence, NaN loss, OOM, low GPU utilisation, failed resume.
Building evaluation sets and analysing model weaknesses such as bias and robustness.
Hindi, Hinglish or other Indic language modelling.
Financial data: bank statements, bureau reports, credit or collections conversations.
Knowledge distillation, quantisation and inference optimisation.
Speech or speech-to-speech models.
Open-source contributions or published checkpoints.
Experience working with business or customer-facing teams.
About Lendenclub
Headquartered in Mumbai, founded in 2015.
- People in Elip's index who work here
- 306
- Typical time at Lendenclub
- 2.2 years
- Typical career length so far
- 4.4 years
Largest teams
- Engineering41 people
- Finance40 people
- Leadership35 people
- Information Technology26 people
Totals from public professional profiles in Elip's index; any group under ten people is left out.
Similar open jobs
About this page
Elip lists jobs that employers publish on their own careers sites, never copies from job boards. This one comes from Lendenclub's listing on Keka, posted 17 Nov 2026. The summary is Elip's; the description is the employer's. When the job closes, this page says so.