Applied Machine Learning Scientist – Voice AI, NLP & GenAI Applications Location : Sector 63, Gurugram, Haryana – 100% In-Office Working Days : Monday to Friday, with 2nd and 4th Saturdays off Working Hours : 10:30 AM – 8:00 PM Experience : 3–7 years in applied ML, with at least 2 years focused on voice, NLP, or GenAI deployments Function : AI/ML Research & Engineering | Conversational Intelligence | Real-time Model Deployment Apply : careers@darwix.ai Subject Line : “Application – Applied ML Scientist – [Your Name]” About Darwix AI Darwix AI is a GenAI-powered platform transforming how enterprise sales, support, and credit teams engage with customers. Our proprietary AI stack ingests data across calls, chat, email, and CCTV streams to generate: Real-time nudges for agents and reps Conversational analytics and scoring to drive performance CCTV-based behavior insights to boost in-store conversion We’re live across leading enterprises in India and MENA, including IndiaMart, Wakefit, Emaar, GIVA, Bank Dofar , and others. We’re backed by top-tier operators and venture investors and scaling rapidly across multiple verticals and geographies. Role Overview We are looking for a hands-on, impact-driven Applied Machine Learning Scientist to build, optimize, and productionize AI models across ASR, NLP, and LLM-driven intelligence layers . This is a core role in our AI/ML team where you’ll be responsible for building the foundational ML capabilities that drive our real-time sales intelligence platform. You will work on large-scale multilingual voice-to-text pipelines, transformer-based intent detection, and retrieval-augmented generation systems used in live enterprise deployments. Key ResponsibilitiesVoice-to-Text (ASR) Engineering Deploy and fine-tune ASR models such as WhisperX, wav2vec 2.0, or DeepSpeech for Indian and GCC languages Integrate diarization and punctuation recovery pipelines Benchmark and improve transcription accuracy across noisy call environments Optimize ASR latency for real-time and batch processing modes NLP & Conversational Intelligence Train and deploy NLP models for sentence classification, intent tagging, sentiment, emotion, and behavioral scoring Build call scoring logic aligned to domain-specific taxonomies (sales pitch, empathy, CTA, etc.) Fine-tune transformers (BERT, RoBERTa, etc.) for multilingual performance Contribute to real-time inference APIs for NLP outputs in live dashboards GenAI & LLM Systems Design and test GenAI prompts for summarization, coaching, and feedback generation Integrate retrieval-augmented generation (RAG) using OpenAI, HuggingFace, or open-source LLMs Collaborate with product and engineering teams to deliver LLM-based features with measurable accuracy and latency metrics Implement prompt tuning, caching, and fallback strategies to ensure system reliability Experimentation & Deployment Own model lifecycle: data preparation, training, evaluation, deployment, monitoring Build reproducible training pipelines using MLflow, DVC, or similar tools Write efficient, well-structured, production-ready code for inference APIs Document experiments and share insights with cross-functional teams Required Qualifications Bachelor’s or Master’s degree in Computer Science, AI, Data Science, or related fields 3–7 years experience applying ML in production, including NLP and/or speech Experience with transformer-based architectures for text or audio (e.g., BERT, Wav2Vec, Whisper) Strong Python skills with experience in PyTorch or TensorFlow Experience with REST APIs, model packaging (FastAPI, Flask, etc.), and containerization (Docker) Familiarity with audio pre-processing, signal enhancement, or feature extraction (MFCC, spectrograms) Knowledge of MLOps tools for experiment tracking, monitoring, and reproducibility Ability to work collaboratively in a fast-paced startup environment Preferred Skills Prior experience working with multilingual datasets (Hindi, Arabic, Tamil, etc.) Knowledge of diarization and speaker separation algorithms Experience with LLM APIs (OpenAI, Cohere, Mistral, LLaMA) and RAG pipelines Familiarity with inference optimization techniques (quantization, ONNX, TorchScript) Contribution to open-source ASR or NLP projects Working knowledge of AWS/GCP/Azure cloud platforms What Success Looks Like Transcription accuracy improvement ≥ 85% across core languages NLP pipelines used in ≥ 80% of Darwix AI’s daily analyzed calls 3–5 LLM-driven product features delivered in the first year Inference latency reduced by 30–50% through model and infra optimization AI features embedded across all Tier 1 customer accounts within 12 months Life at Darwix AI You will be working in a high-velocity product organization where AI is core to our value proposition. You’ll collaborate directly with the founding team and cross-functional leads, have access to enterprise datasets, and work on ML systems that impact large-scale, real-time operations. We value rigor, ownership, and speed. Model ideas become experiments in days, and successful experiments become deployed product features in weeks. Compensation & Perks Competitive fixed salary based on experience Quarterly/Annual performance-linked bonuses ESOP eligibility post 12 months Compute credits and model experimentation environment Health insurance, mental wellness stipend Premium tools and GPU access for model development Learning wallet for certifications, courses, and AI research access Career Path Year 1: Deliver production-grade ASR/NLP/LLM systems for high-usage product modules Year 2: Transition into Senior Applied Scientist or Tech Lead for conversation intelligence Year 3: Grow into Head of Applied AI or Architect-level roles across vertical product lines How to Apply Email the following to careers@darwix.ai : Updated resume (PDF) A short write-up (200 words max): “How would you design and optimize a multilingual voice-to-text and NLP pipeline for noisy call center data in Hindi and English?” Optional: GitHub or portfolio links demonstrating your work Subject Line : “Application – Applied Machine Learning Scientist – [Your Name]” Show more Show less