- Building and owning a conversational AI research assistant for retail investors in Indian capital markets, think Claude but specialized for different instruments and macro analysis for non-expert users

- Designing multi-agent agentic pipelines with tool-calling, memory management, and multi-turn conversational flows that handle real, messy, incomplete queries from retail users (not just clean, structured prompts from analysts)

- RAG pipeline architecture : source curation, chunking strategy, embedding quality, retrieval tuning, reranking, and citation-grounded responses that retail users can trust

- Integrating real-time financial data sources (NSE/BSE feeds, Screener, Tickertape, news APIs, company filings) as live tool-callable data layers, not just static retrieval

- Building the guardrails and evaluation layer : domain scoping, hallucination mitigation, confidence scoring, and monitoring to ensure the system stays accurate and within bounds over time

- Fine-tuning or adapting LLMs where retrieval alone isn't sufficient

- Building ML models for user behaviour, personalization, and financial insights that feed into the conversational layer

Expectations :

- Understanding of large language models (LLMs) like LLAMA, Anthropic Claude 3, or Sonnet.

- Familiarity with cloud platforms for data science like AWS Bedrock and GCP Vertex AI

- Strong proficiency in Python and data science libraries (scikit-learn, TensorFlow, PyTorch).

- Solid understanding of statistical methods, machine learning algorithms, and wealth tech applications.

- Experience in data wrangling, visualization, and analysis.

- Collaborative mindset and ability to thrive in a fast-paced startup environment.

Bonus points :

- Experience in capital market usecases

- Familiarity with recommender systems and personalization techniques.