Python Developer

About Sandhata

Sandhata Technologies is a 20-year

Role overview

Join a small, hands-on team building an Agentic AI

What you’ll do

• Build backend services in 

  Python + FastAPI

   to orchestrate 

  LLM

   (large language model) agents and tools.
• Implement agent workflows (planning, tool use, retries, fallbacks), using frameworks like 

  LangGraph/LangChain/LlamaIndex

   (or equivalents).
• Design and tune 

  RAG

   (retrieval-augmented generation): ingestion pipelines, chunking, embeddings, and 

  vector DB

   (e.g., 

  pgvector

  , Pinecone, Weaviate).
• Integrate enterprise/banking systems (KYC, payments, case management) via REST/gRPC; handle 

  idempotency

  , 

  rate limiting

  , backoff, and circuit breakers.
• Add 

  guardrails

   (policy checks, PII—personally identifiable information—redaction, prompt templates, function-calling schemas), and observability (structured logging, tracing, metrics).
• Build evaluation loops and offline tests for AI quality (e.g., 

  Ragas/DeepEval

  ), and production monitoring (

  Langfuse/Phoenix

  ) for drift, latency, and cost.
• Ensure security/compliance: authN/Z (OAuth2/JWT, 

  RBAC

  —role-based access control), encryption, audit logs, data retention, and least-privilege access.
• Operate services with 

  Docker

  , CI/CD, and basic cloud (AWS/Azure/GCP); collaborate on infra (K8s/Terraform exposure is a plus).
• Own features end-to-end—requirements → design → code → test → deploy → monitor.

Must-have

• 4–6 years

   building backend systems in 

  Python

   (production use), with 

  FastAPI

   (or Django + DRF) and REST API design.
• Practical LLM integration experience (OpenAI/Azure OpenAI/Anthropic/Bedrock or similar), including function calling/tool use.
• Hands-on with 

  RAG

   and a 

  vector database

   (pgvector/Pinecone/Weaviate) and embeddings (e.g., sentence-transformers).
• Strong 

  SQL

   (PostgreSQL/MySQL): schema design, indexes, transactions; plus 

  Redis

   for caching/queues; 

  Celery

  or async tasks.
• Solid engineering foundations: testing (

  pytest

  ), error handling, retries, observability, performance tuning.
• Comfort working 

  independently

   in a fast, regulated setting; clear written/verbal communication.

Good to have

• Agent orchestration frameworks (

  LangGraph/LangChain/LlamaIndex

  ) and 

  LLMOps

   (prompt/versioning, cost controls).
• Event-driven patterns (

  Kafka/RabbitMQ

  ), streaming, and batch pipelines.
• Model hosting/serving (vLLM/TGI) and EVA (evaluation) design.
• Security/compliance exposure in 

  banking/fintech

   (audit trails, data residency, SOC2/PCI considerations).
• Kubernetes/Terraform, GitHub Actions/Azure DevOps CI/CD.

Work model & location

• On-site

   at Sandhata’s 

  Chennai

   office (Mon–Fri). Chennai-based candidates preferred; relocators welcome.

Why Sandhata

• Ship real 

  Agentic AI

   capabilities for a Tier-1 banking use case.
• Modern stack, ownership culture, supportive leadership, and learning budget.

Compensation

• Competitive; aligned to experience and impact.