About Swaran Soft
Swaran Soft Support Solutions Pvt. Ltd. is an enterprise AI and digital transformation company headquartered in Gurugram, India, with offices in the UAE, Estonia, and the USA. Our Agentic AI Consulting & Managed Services division builds intelligent automation platforms for large Indian and Middle Eastern enterprises across BFSI, telecom, healthcare, and manufacturing sectors.
Our platform is built on open-source infrastructure — N8N for workflow orchestration, Supabase for secure backends, and integrated LLMs for real-time voice and text intelligence. We support 9+ Indian languages natively and work with both global LLMs and Indian sovereign AI models including Sarvam AI, BharatGen, and Krutrim. Every Agentic AI developer at Swaran Soft builds systems that run in production at enterprise scale — not demos, not prototypes.
explore our work, our technology philosophy, and what it looks like to build
Agentic AI for India's most complex enterprises.
The Role
As an Agentic AI Developer at Swaran Soft, you will design, build, and deploy intelligent agent systems that automate complex enterprise workflows. You will work across the full stack of agentic AI development — from data acquisition through web scraping and crawling, to agent orchestration using LangChain and LangGraph, to deployment on enterprise platforms via N8N and Supabase.
You will not be maintaining legacy systems or building internal tools. You will be building client-facing AI systems that handle real enterprise processes — customer support automation, multilingual voice agents, intelligent ticketing, sentiment-driven escalation, and data extraction pipelines. Your work will be seen, used, and evaluated by enterprise stakeholders from day one.
What You Will Build & Own
Agentic AI Systems & Orchestration
- Design and build multi-agent AI systems using LangChain and LangGraph — including agent memory, tool use, planning loops, and inter-agent communication
- Develop and deploy intelligent workflow automations using N8N — covering enterprise integrations across WhatsApp, Microsoft Teams, CRM systems (Zoho, Freshdesk, ServiceNow), and telephony APIs (Exotel, Twilio)
- Build conversational AI agents and chatbots with context retention, intent recognition, and multi-turn dialogue management
- Implement sentiment analysis, intent detection, and tone classification pipelines for voice and text data using LLM-native NLP
- Create human-in-the-loop escalation systems — agents that detect resolution limits and hand off to human operators with context summaries
Data Acquisition & Extraction Pipelines
- Build robust web scraping and crawling systems to acquire structured and unstructured data from enterprise and public web sources
- Develop data extraction pipelines that clean, normalise, and route scraped data into vector databases and Supabase backends
- Design and implement RAG (Retrieval-Augmented Generation) pipelines — connecting vector databases to LLMs for grounded, knowledge-accurate agent responses
- Build and maintain vector database schemas (Pinecone, Weaviate, pgvector on Supabase) for semantic search and retrieval at enterprise scale
- Ensure data pipeline reliability, rate-limit handling, error recovery, and structured logging across all scraping and extraction workloads
LLM Engineering & Fine-Tuning
- Fine-tune open-source LLMs (Mistral, LLaMA, or equivalent) on client-specific datasets for domain-adapted performance
- Evaluate and select appropriate LLMs for each client use case — balancing cost, latency, language support, and accuracy requirements
- Integrate and route between hosted LLMs (OpenAI, Mistral via OpenRouter) and locally deployed models via Ollama
- Implement and optimise prompt engineering strategies — chain-of-thought, few-shot, system prompts — for production agent systems
- Work with Indian sovereign AI models (Sarvam AI, BharatGen, Krutrim) for multilingual Indian language deployments
Backend, Infrastructure & Integration
- Build and manage Supabase backends — including real-time PostgreSQL schemas, Row-Level Security policies, edge functions, and authentication
- Design API layers that connect agentic AI systems to enterprise CRM, ITSM, messaging, and telephony platforms
- Deploy and manage AI systems in Docker and Kubernetes environments for cloud, hybrid cloud, and on-premise enterprise clients
- Maintain RBAC and user management configurations — ensuring enterprise-grade access control on all deployed systems
- Write clean, documented, and testable Python code — following production engineering standards, not research or notebook conventions
Skills & Experience
Must-Have — You Will Be Assessed on These
- 2–4 years of hands-on Python development — production-quality code, not just scripts or Jupyter notebooks
- Demonstrated experience building agentic AI systems using LangChain — agents with tool use, memory, and multi-step reasoning
- Working knowledge of LangGraph for stateful, graph-based agent orchestration — you have built at least one production or near-production workflow
- Experience with web scraping and crawling frameworks — BeautifulSoup, Scrapy, Playwright, Selenium, or equivalent — including handling anti-scraping measures, pagination, and dynamic content
- Hands-on experience with Supabase — schema design, RLS policies, Postgres functions, and real-time subscriptions
- Experience with vector databases — pgvector, Pinecone, Weaviate, Chroma, or equivalent — for semantic search and RAG implementation
- Experience building chatbots or conversational agents with multi-turn context management and intent handling
- Working knowledge of N8N for workflow automation — you have built at least one multi-step automation with API integrations
- Understanding of LLM fundamentals — tokenisation, context windows, embedding models, prompt engineering, and inference trade-offs
- Experience with sentiment analysis pipelines — either using pre-trained models or LLM-based classification
Strong Advantage — Sets You Apart
- Hands-on experience with LLM fine-tuning — LoRA, QLoRA, or full fine-tuning on open-source models (Mistral, LLaMA, Phi)
- Experience integrating voice AI stacks — STT (Whisper) and TTS (ElevenLabs, PlayHT) in production agent systems
- Exposure to Indian language NLP — models supporting Hindi, Tamil, Telugu, or other Indian languages
- Experience with enterprise API integrations — WhatsApp Cloud API, Microsoft Teams Bot Framework, Zoho, Freshdesk, Exotel, or Twilio
- Familiarity with Docker and Kubernetes for deploying AI applications in cloud or on-premise environments
- Contributions to open-source AI projects, published technical writing, or demonstrated personal AI projects on GitHub
A Typical Week in This Role
Monday: Sprint planning with the AI team — reviewing the current pilot client's agent architecture, identifying blockers, and scoping the week's deliverables.
Tuesday–Wednesday: Core build time — writing LangGraph agent workflows, wiring N8N automations to enterprise APIs, building a data extraction pipeline for a new client dataset, or implementing a RAG pipeline on a Supabase vector store.
Thursday: Code review, testing, and deployment — validating agent behaviour across edge cases, reviewing a junior engineer's scraping module, and deploying to staging for client review.
Friday: Pre-sales technical support — joining a CTO meeting with a prospective enterprise client to walk through the technical architecture of a proposed Agentic AI pilot, or preparing the technical component of a proposal scope document.
No two weeks are the same. The client roster spans BFSI, telecom, and healthcare — the problems are genuinely different, and the solutions require you to think, not just implement.
What We Offer
- Work on live enterprise AI deployments — not internal tools, not tutorials, not proof-of-concepts that never ship
- Early access to Indian sovereign AI models (Sarvam AI, BharatGen, Krutrim) — a first-mover advantage in the market
- Direct mentorship from senior AI engineers and the CTO in a small, high-ownership team
- Competitive salary benchmarked to the Delhi NCR Agentic AI developer market for 2–4 years experience
- Learning budget for certifications, AI conferences, and open-source contribution time
- Hybrid-friendly working model from Gurugram HQ — flexibility without sacrificing team collaboration
- Clear growth path: Senior AI Engineer → AI Solutions Architect → Practice Lead within 24–36 months based on performance
- A team that genuinely debates LLM selection, agent architecture trade-offs, and prompt strategies — not a team that just ships whatever works
Life at Swaran Soft
We are a small team building large things. Every engineer at Swaran Soft has direct visibility into client outcomes — when an agent we built reduces a client's support resolution time by 40%, everyone on the team knows it and owns it.
We do not have bureaucracy between an idea and its implementation. If you spot a better way to architect an agent workflow, you raise it, we debate it, and if it holds up, we build it. That is the culture.
If you are 2–4 years into an AI
career and you want to work on genuinely hard problems with a team that