Software Engineer

Job Summary

Make Your Mark: We are looking for a deeply hands-on AI Engineer with proven expertise in Agentic AI and Generative AI systems. The ideal candidate has independently driven 3-4 end-to-end AI implementations - from requirements gathering and solution design through to production deployment - and is equally comfortable in writing production-grade code. This role demands both breadth and depth: you will build and orchestrate intelligent agents, and embed robust governance frameworks that make those agents enterprise-ready.

Key Responsibilities 1. End-to-End AI Implementation

• Own 3-4 complete AI delivery cycles: discovery, requirements, solution design, build, test, and production deployment.
• Translate ambiguous business problems into concrete AI solutions with clear success metrics.
• Define and manage delivery milestones across cross-functional stakeholders.
• Conduct design reviews, code reviews, and production readiness assessments.

1. Architecture & System Design

• Design multi-agent pipelines with clear separation of concerns, fault isolation, and observability.
• Define data flows, integration patterns (REST, gRPC, event-driven), and storage strategies for AI workloads.
• Present technical proposals to senior members -technical leads, Staff/Principal Engineers .

1. Agentic AI - Build & Orchestration

• Design, build, and deploy intelligent agents for complex, multi-step reasoning and task execution.
• Implement agent orchestration using Google Workflows, Temporal, LangChain, LangGraph, CrewAI, or equivalent frameworks.
• Deploy agents on GCP (Vertex AI Agent Builder, Cloud Run, GKE) or AWS / Azure equivalents.
• Integrate agents with external tools, APIs, databases, and enterprise systems via function calling and tool-use patterns.
• Implement ReAct, Plan-and-Execute, and other advanced prompting and reasoning strategies.

1. Agent Governance & Enterprise Readiness

• Design immutable, structured audit logs capturing agent decisions, tool invocations, and data accessed - enabling full traceability.
• Implement content safety policies, permission scopes, and fine-grained access controls governing agent behaviour.
• Architect HITL checkpoints for high-risk decisions, escalation flows, and approval gates.
• Human-in-the-Loop (HITL):
• Design secure, reliable A2A protocols including message schemas, authentication, and failure handling.
• Agent-to-Agent (A2A) Communication:
• Establish agent identity frameworks (service accounts, OAuth, signed tokens) for secure system interactions.
• Build or adopt agent SDKs; maintain a centralised agent registry for discovery, versioning, and lifecycle management.
• Agent SDK & Registry:
• Define and instrument evaluation frameworks covering task success rate, latency, token cost, and output quality.
• Evaluation Metrics:
• Implement runtime monitoring for unexpected agent behaviour, runaway loops, and out-of-distribution outputs.
• Anomaly Detection:
• Develop risk scoring models to dynamically assess and throttle agent actions based on context and potential impact.
• Risk Scoring:

1. Generative AI Engineering

• Design and implement RAG (Retrieval-Augmented Generation) pipelines with vector stores (Pinecone, Weaviate, pgvector, etc.).
• Fine-tune, prompt-engineer, and evaluate foundation models (Gemini, GPT-4, Claude, Llama) for domain-specific tasks.
• Optimise inference costs through caching, batching, model routing, and quantisation strategies.
• Build evaluation harnesses to benchmark model quality, factuality, and safety at scale.

Required Qualifications Experience

• 5-6 years of software engineering experience, with at least 4 years focused on AI/ML systems.
• Proven experience designing and deploying complex, production-grade systems at scale.

Technical Skills

• Agentic AI: LangChain, LangGraph, CrewAI, AutoGen, Google Agent Development Kit (ADK), or similar.
• Cloud Platforms: GCP (Vertex AI, Cloud Run, GKE, Workflows) - AWS or Azure equivalents acceptable.
• Orchestration: Google Workflows, Apache Airflow, Prefect, or equivalent.
• LLM APIs: Gemini, OpenAI, Anthropic Claude, Mistral, or open-source models (Llama, Mistral).
• Vector DBs & RAG: Pinecone, Weaviate, Qdrant, pgvector, or Elasticsearch.
• Languages: Python (primary); Go, Java, C#, TypeScript a plus.
• Infra & DevOps: Docker, Kubernetes, Terraform, CI/CD pipelines (GitHub Actions, Cloud Build).
• Observability: OpenTelemetry, Prometheus, Grafana, or equivalent for AI workload monit…