AI Engineer

Andpayments; India; On-site; INR 1,800,000 - 3,000,000; Full time

Job summary

AndPayments is hiring an AI Engineer to design, build, and productionise agentic AI systems powering our analytics and automation platforms. You will own the full lifecycle from problem framing to deployment, evaluation, and monitoring in production.

You will work at the intersection of LLM-based workflows, classical ML, and data engineering, partnering with product and client delivery teams to ship features used by enterprise users.

Qualifications

• Proficiency in Python and building production-grade ML/AI systems.
• Experience with LLM-based or agentic systems in production preferred.
• Strong understanding of data pipelines, ML model lifecycles, and evaluation.

Responsibilities

• Design, build, and productionise agentic AI systems powering analytics and automation.
• Own full lifecycle from problem framing to deployment and monitoring.
• Collaborate with product, backend, and client delivery teams to ship enterprise features.
• Develop multi-agent workflows and robust tool calling interfaces.
• Implement RAG pipelines, prompts, and evaluation harnesses.
• Ensure guardrails, privacy, and cost controls in AI systems.
• Mentor engineers and contribute to internal standards.

Skills

Python

FastAPI

SQL

LangGraph

LangChain

Tool schemas

JSON outputs / tool-calling

Pydantic AI

OpenAI Agents SDK

PyTorch

Time-series forecasting

Statistics

PII detection

Prompt-injection mitigation

Bias and explainability

PostgreSQL

MongoDB

pgvector / Vector databases

NoSQL & relational data modeling

scikit-learn

Education

Bachelor's or Master's in Computer Science, Data Science, Engineering, Mathematics, Statistics

Tools

PostgreSQL

MongoDB

pgvector

Vector databases (Pinecone / Chroma / Milvus)

Docker

Kubernetes

CI/CD

FastAPI

PyTorch

Job description

AndPayments is building the next-generation payments stack for Indian businesses, and for businesses globally that need India-grade payments infrastructure. We operate across Payments, RegTech, B2B SaaS and Prepaid Instruments. These are not separate bets. They are a deliberate stack: compliance infrastructure that enables payment products that enable financial instruments, all built together rather than bolted together.

We move fast because we have to. We care deeply about compliance because we build on regulated infrastructure, and because we believe compliance done right is a competitive moat, not just a cost of doing business. We use AI as a primary operating layer, not as a productivity add-on. If you want to spend the next few years building something that will define how Indian businesses transact, and eventually how businesses transact globally, this is where you should be.

About the role

We are looking for an AI Engineer to design, build, and productionise agentic AI systems and machine learning solutions that power our analytics and automation platforms. This is a hands-on engineering role: you will own the full lifecycle — from problem framing and model/agent design through to deployment, evaluation, and ongoing monitoring in production.

You will work at the intersection of LLM-based agentic workflows, classical machine learning, and data engineering, partnering closely with product, backend, and client delivery teams to ship features that go live with real enterprise users.

What you'll do

Agentic AI & LLM Systems

• Design and build multi-agent and single-agent workflows using LangGraph — including state graphs, conditional routing, checkpointing, memory, and human-in-the-loop interrupts.
• Build and maintain MCP (Model Context Protocol) servers and clients to expose internal tools, data sources, and APIs to LLM agents in a standardised, secure way.
• Design robust tool/function-calling interfaces: schema design, argument validation, error handling, retries, and graceful degradation.
• Implement RAG pipelines end to end — ingestion, chunking, embedding, hybrid retrieval, reranking, citation, and grounding.
• Engineer and version prompts systematically; build structured-output pipelines with schema enforcement.
• Build evaluation harnesses for agents and LLM features (golden datasets, LLM-as-judge, regression suites) and drive measurable quality improvements.
• Implement guardrails: prompt-injection defence, PII redaction, output validation, cost and token budgeting, rate limiting.
• Build, train, and evaluate ML models for classification, regression, forecasting, anomaly detection, and clustering on structured enterprise data.
• Own feature engineering, model selection, hyperparameter tuning, and rigorous offline/online evaluation.
• Fine-tune or adapt open-source LLMs where warranted (LoRA / QLoRA / PEFT, instruction tuning, embedding model fine-tuning).
• Monitor deployed models for drift, degradation, and data-quality issues; own retraining cycles.
• Write production-grade Python: clean, typed, tested, and reviewed.
• Build and expose AI services as APIs (FastAPI), with async handling, streaming responses, and queueing where needed.
• Containerise and deploy workloads to cloud with CI/CD, environment management, and infrastructure-as-code.
• Instrument systems with tracing and observability for latency, cost, token usage, tool-call success rates, and failure modes.
• Translate ambiguous business or client requirements into concrete AI system designs with clear scope and success criteria.
• Write clear technical documentation, architecture notes, and handover material.
• Support UAT cycles, client demos, and production issue triage.
• Mentor junior engineers and contribute to internal engineering standards.

What we're looking for

• Python (expert) - 3+ years; async/await, typing, packaging, testing (pytest)
• FastAPI or equivalent for API development
• SQL - complex queries, joins, window functions, query optimisation
• Git, code review workflows, branching strategies
• LangGraph - required; production experience building stateful, cyclic agent graphs
• LangChain / LlamaIndex - chains, retrievers, document loaders, output parsers
• Exposure to at least one other orchestration framework: CrewAI, AutoGen, OpenAI Agents SDK, Pydantic AI, or Semantic Kernel
• Building MCP servers (tools, resources, prompts) and integrating MCP clients
• Transport handling (stdio, SSE, streamable HTTP), authentication, and tool-scoping
• Designing safe, well-described tool schemas for reliable LLM invocation
• Working with commercial and open models: Claude (Anthropic), GPT (OpenAI), Gemini, Llama, Mistral, Qwen
• Prompt engineering, few-shot design, structured outputs, JSON-mode / tool-calling
• Context management: caching, compression, summarisation, long-context handling
• PostgreSQL - schema design, indexing, joins, window functions, query optimisation; familiarity with the pgvector extension
• MongoDB - document modelling, aggregation pipelines, indexing strategies, and knowing when a document store beats a relational one
• Vector databases: pgvector, Qdrant, Pinecone, Chroma, Weaviate, or Milvus - collection design, metadata filtering, and index tuning
• Comfort working across both relational and NoSQL stores in the same system, including data modelling trade-offs
• scikit-learn, XGBoost / LightGBM, pandas, NumPy
• PyTorch (or TensorFlow) for deep learning and fine-tuning workflows
• Time-series forecasting, imbalanced-data handling, cross-validation, model interpretability (SHAP)
• Solid grounding in statistics and evaluation metrics — knowing which metric matters for the business problem
• PII detection and redaction; data residency and compliance awareness
• Prompt-injection and jailbreak mitigation; least-privilege tool access
• Bias, fairness, and explainability considerations in model deployment
• Bachelor's or Master's degree in Computer Science, Data Science, Engineering, Mathematics, Statistics, or a related field
• 3-4 years of professional experience in AI/ML engineering, with at least 1 year building LLM-based or agentic systems in production
• Demonstrable track record of shipping AI systems that reached real users — not only prototypes or notebooks

Nice to have

• Experience deploying AI systems in BFSI / fintech / regulated environments
• Regulatory or compliance reporting automation experience
• Conversational AI / chatbot systems with multi-turn state management
• Knowledge graphs (Neo4j) or GraphRAG
• Frontend familiarity (React / Streamlit / Gradio) for building internal AI tooling
• Open-source contributions, published work, or an active technical portfolio