AI/ML Engineer

Responsibilities

Design and build agentic AI systems — tool use, multi-agent orchestration, ReAct/chain-of-thought pipelines

Develop and deploy LLM-powered features: RAG pipelines, autonomous workflow automation

Own prompt engineering and context engineering — manage context windows, token budgets, and output structuring

Build and maintain backend services using Python and FastAPI; design and manage relational databases with SQLAlchemy and Alembic migrations

Host and serve custom models on AWS SageMaker; manage endpoints, scaling, and inference pipelines

Work with open-source models via Hugging Face — model selection, inference, and lightweight customisation

Set up and maintain ML/AI-Ops workflows — CI/CD pipelines for model deployment, automated testing, and continuous delivery of AI features

Stay updated with the latest advancements in ML, AI, deep learning, and agentic frameworks

Must have

3+ years of experience in Python (3+) — clean, production-grade code with async patterns and FastAPI

GenAI fundamentals — solid understanding of tokens, embeddings, prompt engineering, and context engineering

Familiarity with Agent-to-Agent (A2A) and Model Context Protocol (MCP)

Basic ML/NLP knowledge — classification, regression, NLP pipelines, BERT, LIWC, Bag of Words, and embedding-based similarity models

Hands-on experience deploying and hosting custom models on AWS SageMaker

Familiarity with Hugging Face ecosystem and open-source model landscape

Experience with ML/AIOps — CI/CD pipelines for model lifecycle, model versioning, monitoring, and automated evaluation in production

Strong grasp of data structures, algorithms, and software engineering principles

Excellent problem-solving skills with ability to work independently and collaboratively in a fast-paced environment

Good written and verbal communication skills

Nice to have

Experience fine-tuning foundation models using PEFT/LoRA techniques

Knowledge of guardrails, safety evaluation, or responsible AI principles

• Monitor models in production — track drift, performance degradation, hallucination rates, token usage, and latency