Job Title

About ZillworkZillwork is a Singapore-incorporated AI company building technology that creates economic opportunity at scale. We are pre-launch, founder-led, and operating out of Chennai and Coimbatore.We are technically ambitious and globally minded. Our stack includes fine-tuned multilingual speech models, real-time geo-intelligent matching at scale, and a homegrown AI engine designed from first principles. We are building for markets and communities that existing technology has never seriously attempted to serve.We are a smart, focused team. Every hire shapes the culture and the product. If you want your work to matter — and you are energised by hard problems over comfort — read on.The RoleAs NLP Lead, you will design and own the complete multilingual NLP architecture at the core of our platform. You will fine-tune speech and language models on real-world informal speech data — code-switching, regional dialects, and vocabulary that generic models consistently fail on.This is the most critical technical hire at Zillwork. You work directly with the founder. Your models ship to production and reach people in ways most AI never does.What You Will Build- Fine-tune speech recognition models (IndicWhisper, Whisper) on informal multilingual speech corpora- Fine-tune NLU models (MuRIL, multilingual transformers) on code-switching and dialect data- Build a structured information extraction pipeline: spoken phrase to structured output- Design the multilingual NLP architecture from voice input to intent resolution- Own model accuracy loop: MLflow experiment tracking, EvidentlyAI drift detection, auto-retraining- Build multilingual semantic embeddings (BGE-M3) for semantic matching at scale- Work on our GPU workstation: 2x NVIDIA RTX 6000 Ada, 96GB ECC VRAMWhat We Are Looking For- 7+ years NLP engineering with production deployment- Hands-on with Indian language models: IndicWhisper, MuRIL, AI4Bharat stack- Tamil language proficiency — native or near-native. Non-negotiable for this role.- Experience fine-tuning transformer models (HuggingFace, PyTorch)- Deep understanding of informal speech: code-switching, dialect variation, low-resource vocabulary- BGE-M3 or multilingual embedding models in production- MLflow or equivalent experiment tracking in productionBonus: Published multilingual NLP research, Sarvam AI or Bhashini experience, additional South Indian language knowledgeApply: hr@ | Subject: NLP Lead Application