AI jobs in fintech and banking in India: the roles, the rules and who is hiring

Why fintech and banking are hiring AI now

Three things changed at once: adoption reached the point where banks need owners for AI, the regulator set out what it expects of them, and payment volumes outgrew manual operations.

The regulator now has a position. The RBI's FREE-AI committee set out 7 Sutras and 26 recommendations in August 2025. Chaired by Dr Pushpak Bhattacharyya of IIT Bombay, it recommends a board-approved AI policy for every regulated entity, a model inventory with validation and ongoing monitoring for drift, human oversight for medium and high-risk use cases, data governance aligned with the Digital Personal Data Protection (DPDP) Act, and a route for customers to reach a human. In June 2026 the RBI followed with a draft Guidance on Regulatory Principles for Model Risk Management that applies to "all models used by regulated entities", including third-party models and models using AI and ML, across banks, NBFCs, cooperative banks and credit information companies. Separately, the DPDP Rules, notified on 14 November 2025, give data fiduciaries an 18-month phased compliance period.

At FIBAC in August 2026, Governor Sanjay Malhotra told bankers that "'The model decided' can never be an acceptable answer" to a customer, auditor or the RBI, and listed his expectations: a complete AI system inventory, board-approved governance, explainability for material decisions, pre-deployment red-teaming and stress testing, and meaningful human oversight (RBI speech). Each expectation needs people to deliver it.

The sector is signalling where it is going. Global Fintech Fest 2026 in Mumbai listed ElevenLabs as its "Voice AI Partner", Sarvam as its "AI Ecosystem Partner" and Razorpay as its "AI Powered Payments Partner" (GFF partners). That billing says what the industry plans to buy, and so whom it plans to hire.

The AI roles in fintech and banking

The eight roles below are the ones Indian financial services employers are building teams around. Each one answers, directly or indirectly, to a risk committee.

Head of AI or Chief AI Officer at a bank or NBFC

State Bank of India's October 2026 notice for a General Manager and Chief AI Officer asks for 15 or more years in banking, BFSI or IT, including at least three years of senior leadership in AI, analytics, model oversight or governance, excluding individual contributor time. The role spans enterprise AI platforms including generative and agentic AI, compliance and oversight of independent model validation. In February 2026, Kotak Mahindra Bank designated a Head of Innovation and AI as Senior Management Personnel in a stock exchange filing.

What makes it different. The bank's AI policy is board-approved under FREE-AI, so this person defends AI decisions to the board and supervisors. Compared with the generic seat in our AI jobs in India guide, the bank version adds model risk, data localisation and conduct risk. How it is assessed: a record of AI systems in production in a regulated setting, a view on which use cases should not be automated, and a credible answer to how an AI decision would be explained to the RBI. See our Head of AI search in Bengaluru.

AI risk and model governance lead (FREE-AI)

This title barely existed two years ago. SBI's notice includes a Deputy General Manager for AI Governance and Compliance, responsible for AI governance policies, risk classification, oversight of model validation, regulatory adherence to the RBI and DPDP, and audits and remediation.

What the work is. Maintaining the model inventory the RBI's 2024 draft on model risk in credit and the 2026 draft both expect, running independent validation, demanding outcomes that are "consistent, unbiased, explainable and verifiable", and assessing third-party and vendor models. FICCI, IBA and BCG found 51 per cent of Indian BFSI firms handle third-party AI risk ad hoc or reactively, the gap this person closes. Skills: model validation for ML and LLMs, bias testing, red-teaming, data protection law and writing a policy a board will approve. Background: credit risk modelling, validation, audit or data privacy, with enough technical depth to challenge an ML team.

ML engineer for credit underwriting

Credit models decide who gets a loan and at what price, making this the most regulated ML job in the sector. Bajaj Finance reported a 20 per cent improvement in underwriter efficiency in Q1 FY27, alongside 32 million application documents processed by AI (Bajaj Finance investor presentation).

What the work is. Scorecards and underwriting models, alternative-data features, document extraction and drift monitoring. Under the RBI's 2024 draft, deploying a credit model or changing its inputs needs approval from the board's Risk Management Committee or a designated sub-committee. Under the Digital Lending Directions, 2025, data collection must be need-based, with prior and explicit consent and an audit trail, and borrower data must sit on servers in India. How it is assessed: a case on a real credit problem (reject inference, thin-file borrowers, a model whose approval rate has drifted), plus how you would document the model for validation. A plain-language decline reason matters as much as AUC.

Fraud and AML data scientist

Fraud work is inline, latency-bound and adversarial. At about 802 million UPI transactions a day, rules alone do not keep up. The RBI itself built MuleHunter.AI through the Reserve Bank Innovation Hub in Bengaluru to detect mule accounts, piloted with two public sector banks as shared infrastructure for banks that cannot build their own. SBI's notice includes a DGM for Risk and Operational Analytics building early warning and fraud models with human-in-the-loop controls and audit trails.

Skills: graph features and network analysis for mule rings, streaming feature pipelines, imbalanced classification, adversarial drift, alert prioritisation, and the AML typologies financial intelligence teams care about. How it is assessed: a case on a fraud pattern that shifts after you block it, and on false positive cost: every blocked genuine payment is a complaint. See our explainer on real-time fraud and risk decisioning.

Voice AI and conversational AI engineer

Voice is where AI reaches Indian borrowers at volume. Mahindra Finance's voice agents, built by Mahindra AI on Sarvam's platform, have made more than one crore calls in 12 Indian languages, covering collections from pre-due reminders to write-off and repossession stages, with every call analysed and connected to CRM and loan servicing systems. At GFF 2026, Yes Bank and Open announced an agentic layer for recurring payments that contacts borrowers by voice and WhatsApp in 22 languages to fix likely debit failures (Inc42). The FICCI, IBA and BCG report estimates AI voice and digital channels can collect at about one-tenth the cost of field agents.

What the work is. Speech tuning for Indian languages and code-mixed speech, dialogue design, latency and loan system integration. What makes it regulated: FREE-AI says customers should know they are dealing with AI and be able to reach a human, and every call is evidence in a grievance. Background: speech, NLP or LLM application engineering.

Agentic AI engineer for operations

Operations is where agents replace queues: reconciliation, KYC review, disputes, mandate failures. Bajaj Finance reported 62 agentic AI agents live in Q1 FY27 and plans more than 600 autonomous agents and 118 agentic use cases in FY27. Axis Bank's AI team describes automating lending checks, fraud reviews and reporting with human validation kept in place (Inc42). NPCI's in-house model FiMI already handles UPI dispute queries and mandate lifecycle questions through the UPI Help Assistant (Business Standard).

Skills: agent orchestration, tool calling against core banking and payment systems, deterministic fallbacks, evaluation on historical cases, and tracing every action. What makes it regulated: an agent that changes a customer record or reverses a transaction needs permissions, logs and a human checkpoint the auditor can see. Knowing KYC and AML engineering often decides it.

Forward deployed engineer at AI vendors selling to BFSI

AI vendors to Indian banks need engineers who can make the product work inside a bank's network. Gnani AI launched a BFSI platform with more than 200 pre-built workflows at GFF 2026; Sarvam and ElevenLabs took headline partner slots. Integrating a vendor's model with a bank's identity, logging and data residency controls is the FDE's job.

What makes it different from a general FDE role: security reviews, on-premises or India-hosted deployments, procurement and vendor risk questionnaires, and evaluation on the bank's own historical cases before go-live. See our forward deployed engineer search.

AI product manager

The AI product manager decides which journeys get automated and how far. Axis Bank's approach is instructive: conversational servicing limited to high-confidence journeys such as card blocking, AI-assisted sales for relationship managers, and fully AI-driven journeys kept experimental because of accuracy, cost and regulatory risk. Use cases are measured net of model and token costs.

Skills: fintech product experience in lending, payments or servicing, evaluation literacy, a working knowledge of FREE-AI and the Digital Lending Directions, and the judgement to stop a launch. How it is assessed: a case on whether to automate a specific journey, what the fallback is, and what metric would make you roll it back.

Who is hiring

Hiring spans five employer types. We name companies only where a public source says they are building or hiring for AI.

Banks. State Bank of India is recruiting six fixed-term AI posts in Mumbai, including an Enterprise AI Architect for an NVIDIA-based on-premises AI platform, with registration open until 9 November 2026. Kotak Mahindra Bank has a Head of Innovation and AI in senior management. Axis Bank's CEO said on the Q4 FY26 call that the bank is building "an AI-led customer-centric bank" through its Axiom operating model and expects meaningful bottom-line impact from AI over 18 to 24 months; the same call reported total workforce down 3 per cent year on year. Banks are adding AI builders while slowing hiring elsewhere.

NBFCs and lenders. Bajaj Finance's dedicated AI unit grew from 107 people in Q1 FY26 to 230 in Q1 FY27, and the company plans to add 400 more in FY27; its FY26 annual report describes a governance framework aligned with FREE-AI and a dedicated AI security team. Mahindra Finance describes its Sarvam voice deployment as one of the largest in Indian financial services.

Payments companies. NPCI is building a payments-native AI foundation model with NVIDIA on top of FiMI. Razorpay, NPCI and OpenAI launched an agentic payments pilot on ChatGPT in October 2025, backed by Axis Bank and Airtel Payments Bank. Open built the agentic recurring payments layer with Yes Bank. Our list of top fintech companies in India covers the wider employer set.

Wealth and broking. AI work here (advisory copilots, surveillance, client servicing) is less publicly documented, so we name no firms.

AI vendors to BFSI. Sarvam, ElevenLabs and Gnani AI are the visible names from GFF 2026; vendors like these need FDEs and voice AI engineers who know banking. Our tracker of funded and hiring companies in India follows the funding side.

Pay

AI pay in financial services follows two markets: general AI engineering, and AI applied to a hard financial problem. The second pays more. We do not add new estimates here.

On our 2027 fintech salary benchmark, a fraud or credit ML engineer sits at ₹45 lakh to ₹80 lakh in annual cash against ₹42 lakh to ₹72 lakh for a general fintech ML engineer, and a Head of Data Science or risk ML at ₹77 lakh to ₹1.21 crore. The same benchmark puts the premium for AI applied to fraud or credit at 20 to 35 per cent over a generic equivalent. For Head of AI, AI engineer and forward deployed engineer estimates across all sectors, see AI jobs in India. All of these are SYTalent planning bands and estimates, not survey data.

How to move into AI in fintech

There are three common routes in, each with a predictable gap.

From fintech engineering. You know payment flows, ledgers and KYC. The gap is ML and evaluation: build one agent on real operational data, such as dispute triage, with a written evaluation, and you are a credible agentic AI engineer or FDE candidate.

From ML elsewhere. The gap is regulation and domain. Learn what FREE-AI, the model risk drafts and the Digital Lending Directions require, and why a credit model needs a decline reason. E-commerce and adtech ML candidates often do well in fraud, where adversarial instincts transfer.

From risk and analytics. You know credit, fraud or validation. The gap is engineering: production pipelines, LLM evaluation, agents. Close it and you are the strongest candidate for AI governance roles, because few people have both halves. The Head of Data Science seat is a common destination.

From our desk

These are qualitative observations from the AI and fintech searches we run, not statistics.

• "How would you explain this to the RBI?" now comes up in first rounds. Candidates who have never defended a model struggle.
• Domain is the tiebreaker. Between two equally strong AI engineers, banks pick the one who has seen a credit file, a chargeback or a KYC rejection.
• Governance candidates are the hardest to find. Risk people rarely know LLMs; ML people rarely know validation.
• Voice AI hiring is outrunning supply. Engineers who have shipped Indian-language voice in production are a small pool.

Live roles

Apply once, to us; only shortlisted candidates are contacted.

• Head of AI, Bengaluru
• AI Engineer
• Forward Deployed Engineers, Bengaluru
• Head of Risk, fintech, Mumbai
• All live jobs

Candidates can join the SYTalent network for unadvertised roles. Employers can read about our AI and agentic AI hiring practice or our fintech talent search in India.

Sources

Every figure on this page comes from the sources below. Pay figures come from SYTalent's published estimates and are not drawn from them.