Role Overview

As an SDE-2, you will be responsible for identifying and implementing cost optimisation opportunities across our AI and video-generation infrastructure. You will work on improving system efficiency, reducing infrastructure and model-inference costs, optimising resource utilisation, and ensuring high performance at scale.

This role requires someone who can go beyond writing code and understand the relationship between architecture, performance, scalability, and infrastructure costs.

What you'll do :

1. Design and build scalable Python backend services and APIs for high-volume video-generation workflows.

2. Own distributed job processing (Celery, RabbitMQ). This covers scheduling, prioritisation, retries, idempotency and failure recovery.

3. Make system-design decisions that balance latency, reliability, output quality and cost, and write them up for the team.

4. Design caching and data flows (Redis, MySQL, object storage) that remove redundant compute and API calls.

5. Work hands-on with infra: write Dockerfiles, run and debug workloads on Kubernetes (AKS/EKS), tune autoscaling, and help keep GPU/CPU utilisation high.

6. Build monitoring and cost visibility, so we know what each pipeline and each video costs.

7. Profile, benchmark and load-test services to find bottlenecks.

8 .Work with the AI, Product and Infra teams on architecture.

Must have :

1. 3-5 years building backend systems, mainly in Python.

2. Strong system design: APIs, data modelling, async processing, queues, caching, and designing for failure.

3. Production experience with a task queue (Celery or similar) and Redis.

4. Comfortable with Docker and Kubernetes basics: deploying, debugging pods, and resource requests and limits.

5. Hands-on experience with at least one cloud (AWS or Azure preferred).

6. Strong debugging skills and ownership of problems end to end.

Good to have :

1. GenAI or ML inference pipelines, GPU workloads, model serving, batching or quantisation.

2. FFmpeg or media processing.

3. Prometheus/Grafana or similar observability tools.

4. Cloud cost tools (AWS Cost Explorer, Azure Cost Management, Kubecost).

5. High-scale B2B SaaS experience.

What you'll own in your first 6 months :

1. Cost and usage visibility for each pipeline.

2. Reliability of the bulk video job pipeline (queueing, retries, scale-down of idle workers).

3. At least one measurable cost or latency win, shipped and tracked.

4. Cost / performance: cost optimisation, cost reduction, Kubecost, Cost Explorer, autoscaling, spot instances, profiling, load testing, latency reduction.