Interesting Opportunity Software Engineer (Delhi)

Description - Design and build low-latency AI/ML models for trading signals, market trend prediction, and fraud detection. - Fine-tune and deploy open-source LLMs (Llama 3, Mistral, Gemma) for market-aware RAG (Retrieval-Augmented Generation) and AI-driven advisory systems. - Architect and maintain serverless AI pipelines on AWS using SageMaker, Lambda, Fargate, and Bedrock. - Build and optimize ETL pipelines for high-frequency financial and tick data using AWS Glue and Athena. - Develop scalable, production-ready GenAI workflows using LangChain, vector databases, and retrieval systems. - Ensure SEBI-compliant, privacy-preserving, and secure ML pipelines aligned with financial regulations. - Collaborate closely with product, trading, compliance, and engineering teams to deliver high-impact AI solutions. - Monitor model performance, latency, drift, and continuously improve system reliability. Required Skills & Qualifications - 26 years of hands-on experience in AI/ML, Data Science, or GenAI engineering roles. - Strong programming skills in Python and SQL. - Hands-on experience with PyTorch, Hugging Face Transformers, and GenAI frameworks. - Practical experience with LLM fine-tuning techniques such as LoRA and QLoRA. - Strong understanding and implementation experience of RAG architectures. - Experience working with vector databases such as Qdrant or Milvus. - Solid understanding of real-time or near real-time data processing systems. - Robust experience with AWS services: SageMaker, Bedrock, Lambda, S3, ECR, Fargate. - Knowledge of financial markets, trading systems, or fintech workflows is highly desirable. Tech Stack - LLMs & GenAI: Llama 3, Mistral, Gemma, LangChain - Fine-Tuning: LoRA, QLoRA - Languages: Python, SQL (Bonus: C++) - ML Frameworks: PyTorch, Transformers - Vector DBs: Qdrant, Milvus - Cloud: AWS (SageMaker, Bedrock, Lambda, S3, Glue, Athena, ECR, Fargate) Nice To Have/Bonus Skills - Experience with C++ and latency optimization for real-time system .