Role Summary We are establishing a specialized AI/ML team in Mumbai to modernize our Regulatory Reporting function. We are looking for a hands-on technical leader to build and deploy two classes of solutions: (1) Anomaly Detection Models to catch data quality issues before they reach the regulators, and (2) GenAI Workflows (LLMs/Agents) to automate manual reconciliation and document reviews.
Crucially, you will solve the "Validation Bottleneck." You will design "Human-in-the-Loop" (HITL) workflows that make it easy for business users to validate model outputs against legal loan documents, bridging the gap between "Black Box" AI and auditable regulatory standards.

---

• Key Responsibilities1. ML for Anomaly Detection (The "Watchdog")
• Build unsupervised and semi-supervised ML models (e.g., Isolation Forests, Autoencoders) to scan millions of transactional records for outliers.
• The Challenge: Go beyond simple "threshold checks." Detect complex patterns (e.g., "This trade structure looks valid in isolation but is anomalous for this specific counterparty type").
• Reduce false positives to ensure the Reporting Team trusts the model alerts.
• 1. GenAI & Workflow Automation (The "Builder")
• Design RAG (Retrieval-Augmented Generation) pipelines to "chat" with unstructured data (Credit Agreements, Loan Docs) and extract key regulatory attributes (Maturity Dates, Collateral Clauses).
• Build "Agentic" workflows where GenAI proactively suggests mapping logic or identifies the root cause of a break, requiring only a "thumbs up/down" from the human SME.
• 1. Solving Model Validation & Governance (The "Diplomat")
• This is a critical success factor. You must build "Explainability" (XAI) into every model. You cannot just output a score; you must output why (e.g., "Flagged because this value is 3x higher than the historical average for this product").
• Create Validation Interfaces: Build simple UIs (using Streamlit or React) where business users can see the Model's Prediction side-by-side with the Source Document to rapidly approve/reject the finding.
• Work with Model Risk Management (MRM) to establish a "fast-track" validation framework for non-deterministic GenAI models.

1. Act as the "AI Evangelist" to the Operations/Finance teams, demonstrating how AI assists them rather than replacing them.

---

Candidate Profile (The "Mumbai Persona")

• 1. Technical "Must-Haves"
• Core ML: 6+ years in Data Science/Engineering. Deep experience with Scikit-learn, Tensor Flow, or Py Torch.
• GenAI Stack: Hands-on experience with LLM orchestration frameworks (Lang Chain, Llama Index) and Vector Databases (Pinecone, Milvus, or pgvector).
• The "Validation" Stack: Experience building tools like Streamlit or Gradio for rapid prototyping of human-review interfaces.
• 1. Domain "Nice-to-Haves"
• Experience in Financial Services (specifically Fraud Detection, AML, or Risk Modeling).
• 1. The "X-Factor"
• Communication: Can they explain "Hallucination Risk" to a non-technical Chief Risk Officer?
• Pragmatism: Knows when not to use AI. (e.g., "We don't need an LLM for this; a Regex script is faster and 100% accurate.")

Job Family Group:

Finance

Job Family:

Regulatory Reporting

Time Type:

Most Relevant Skills

Business Acumen, Change Management, Communication, Data Analysis, Financial Acumen, Internal Controls, Issue Management, Problem Solving, Regulatory Reporting.

Other Relevant Skills

For complementary skills, please see above and/or contact the recruiter.

Citi is an equal opportunity employer, and qualified candidates will receive consideration without regard to their race, color, religion, sex, sexual orientation, gender identity, national origin, disability, status as a protected veteran, or any other characteristic protected by law.

If you are a person with a disability and need a reasonable accommodation to use our search tools and/or apply for a career opportunity review Accessibility at Citi.

View Citi’s EEO Policy Statement and the Know Your Rights poster.