Generative AI Application Developer

Job description

Location: Remote (with Bangalore Data Science team)

Experience Required: 5+ years

Contract Type: Full-time

Salary - ₹18-28 LPA dependent on experience

About Zavmo

Zavmo, your lifelong learning companion, is an AI-powered learning platform transforming professional development through intelligent personalisation.

We've built an 18-agent multi-agent system that delivers the 4D Process (Discover, Discuss, Deliver, Demonstrate), integrating National Occupational Standards, OFQUAL frameworks, and career progression intelligence.

Recently recognised as Top 20 AI Innovation Learning Tools by Training Magazine (2025).

The Role

We need an experienced AI engineer who can build and enhance specialised agents within our production multi-agent architecture.

You'll be working with a 4-person Data Science team in Bangalore, implementing RAG-based agents that handle everything from role matching and career progression to assessment and quality assurance.

This isn't a research role - we have clear specifications.

You'll be shipping production code that serves real learners pursuing regulated qualifications.

What You'll Build

Immediate Projects:

• Agent #3: Career Progression Intelligence Agent (RAG + pathway ranking algorithms)
• Agent #16: Learning Specification Triangulation Agent (LO/LO/AC alignment validation)
• Agent #17: Assessor Agent (formative/summative assessment with PASS/REFER/FAIL logic)
• Agent #18: Quality Assurance Agent (meta-agent for self-critique across all agents)

Enhancement Work:

• Expand Agent #1 (Role Matching) to handle 5,000+ job descriptions across 23 sectors
• Enhance Agent #8 (Content Recommendation) with job description contextualisation
• Integrate Agent #10 (Conversational Facilitator) with role-specific examples
  

Core Technical Stack (Must Have 5+ Years)

Primary Technologies:

• Python - All agent logic, RAG implementation, API development
• Vector Databases - Pinecone, Weaviate, Qdrant, or ChromaDB (operational production experience)
• LLM Integration - Anthropic Claude API, prompt engineering, structured outputs
• RAG Architecture - Retrieval Augmented Generation, semantic search, hybrid search
• Embedding Models - OpenAI, Cohere, or Sentence Transformers (generation and vector operations)

Essential Skills:

• API design and RESTful services
• JSON/data structure design
• Database schema design and migrations
• Integration testing and validation
• Error handling and system reliability
  

Required Experience

You Must Have:

• 5+ years Python development in production environments
• 2+ years working with vector databases in production
• 2+ years building RAG-based systems or LLM applications
• Proven experience with embedding generation and semantic search
• Experience designing and implementing APIs for agent-based systems
• Strong understanding of prompt engineering and LLM behaviour

Ideally You Also Have:

• Experience with multi-agent systems or agent orchestration
• Background in ML classification (scikit-learn, PyTorch)
• Knowledge of hybrid search algorithms (vector + keyword)
• Experience with xAPI or learning analytics standards
• Understanding of UK education frameworks (NOS, OFQUAL) - nice to have
• HR/L&D domain knowledge - nice to have
  

What We're Not Looking For

This role isn't suitable if you:

• Need extensive training on Python, vectors, or RAG fundamentals
• Haven't worked with production LLM applications before
• Don't have hands-on vector database experience
• Primarily work in research rather than production systems
• Need close supervision to implement clearly specified features
  

The Team

You'll work with:

• 4 Data Scientists in Bangalore - Your immediate team (DS1-DS4 structure)
• Juliette Denny - Founder, architect of the 4D Process and agent specifications
• Cosmo (CTO) - ADHD brain like a solar system, technical architecture lead
• Gerard (Chairman) - Strategic oversight
  

How We Work

Development Approach:

• Clear specifications with detailed architecture documents
• ReACT pattern (Reasoning → Action → Critique) for all agents
• Work packages with defined deliverables and acceptance criteria
• Parallel workstreams across the 4-person DS team
• Regular integration testing between agents

Current Sprint Example:

• B1: Agent #3 Core Logic (DS3, 7 days)
• B2: Integration Layer + Agent #16, #18 (DS4, 5 days)
• B3: Enhanced Agent #1 (DS3, 3 days)
• B4: Enhanced Agent #8 + #10 (DS4, 3 days)

Timeline: 3-4 week sprints, production deployments

Technical Environment

Data Infrastructure:

• 5,000+ vectorised job descriptions across 23 sectors
• NOS Database (vectorised)
• OFQUAL Register (vectorised)
• Career Pathways Database (500+ progression routes)
• User Profile Database (operational)

Performance Requirements:

• Response times <
  3 seconds
• 95%+ accuracy on role matching
• Semantic consistency across all vector operations
• Complete audit trails for OFQUAL compliance
  

Why This Role Matters

You'll be building AI that genuinely helps people progress in their careers.

Not chatbots that hallucinate.

Not marketing demos.

Real production systems that:

• Match learners to 5,000+ career roles with 95% accuracy
• Map progression pathways with skill gap analysis
• Deliver regulated qualifications (OFQUAL-compliant)
• Provide formal assessments with human oversight

Your code will serve hundreds of thousands of learners globally, helping them live their ideal life design.

What We Offer

• Competitive salary (based on experience)
• Remote-first culture
• Work with cutting-edge AI technology in production
• Direct impact on people's career progression
• Collaborative team that ships real products
• Opportunity to work across the full agent architecture

Required Skill Profession

Computer Occupations