I provide optimized solutions and production-ready pipelines in Generative AI (GENAI), Machine Learning, Computer Vision, and video analytics. My work focuses on measurable system improvements: reduced inference time, lower resource consumption, predictable behavior, and clear operational handover for engineering teams.

• Production-ready AI systems and pipelines for GENAI, ML, Computer Vision, and video analytics optimized for latency, throughput, and cost.
• Reproducible workflows: data ingestion → preprocessing → training → evaluation → inference → monitoring.
• Deployable services (APIs, containers, orchestration) tuned for target workloads and failure modes.
• Documentation, operational runbooks, and handover artefacts for engineering teams.

• Computer Vision

  • Object detection, instance segmentation, multi-camera tracking, and action recognition.
  • Video analytics (surveillance use cases) implemented as part of computer vision systems: stream ingestion, frame sampling, detection → tracking → alerting, with configurable privacy and data‑retention controls.
  • Production considerations: model pruning and quantization, batching, asynchronous processing, and efficient use of accelerators to meet latency and resource objectives.
  • Operational notes: clear evaluation metrics, documented failure modes, and runbooks so teams can operate and extend the system.

• Generative & Multimodal AI

  • Image captioning, multimodal retrieval, prompt-guided generation, and integration of GENAI components into pipelines with attention to latency and cost.
  • Designs that balance model capability with operational constraints and expected workload.

• Machine Learning Engineering & Optimization

  • Data validation, feature pipelines, model selection, and hyperparameter tuning focused on production applicability.
  • Performance profiling and targeted optimizations for I/O, data pipelines, and model serving.

• MLOps & Deployment

  • Containerized APIs (FastAPI), message brokers (RabbitMQ/Kafka), and orchestration with Docker and Kubernetes.
  • CI/CD for models and infrastructure, model versioning, staged rollouts, monitoring, and alerting.
  • Resource and cost planning for multi-tenant or multi-client environments.

• Agentic Systems

  • Design of agents that execute task sequences, handle expected failure modes, and apply model outputs within defined safeguards.

1. Define objectives and KPIs (latency, throughput, accuracy, cost).
2. Prototype and measure baseline performance.
3. Optimize models and pipelines (compression, batching, I/O improvements).
4. Harden production behaviors (logging, retries, isolation, access controls).
5. Deploy and operate with monitoring and model lifecycle management.

• Systems designed to process live streams or stored video at scale, balancing detection rate, false positives, and compute cost.
• Common optimizations:
  • Frame skipping and adaptive sampling based on scene dynamics.
  • Edge preprocessing to reduce bandwidth and cloud compute.
  • Quantized or lower-precision models for edge inference.
  • Partitioning pipeline stages across devices and services to reduce end-to-end latency.
• Deployments include privacy-by-design considerations and configurable retention policies.

• Languages: Python
• ML / DL: PyTorch, ONNX, TensorRT
• Serving & APIs: FastAPI, Uvicorn
• Containers & Orchestration: Docker, Kubernetes
• Messaging & Streams: RabbitMQ, Kafka
• CI / Infra: GitHub Actions, Terraform
• Observability: Prometheus, Grafana, ELK

• Email: aliyannew16@gmail.com
• Website: https://aaliyanahmed-rao.vercel.app/
• GitHub: https://github.com/aaliyanahmed1
• LinkedIn: https://www.linkedin.com/in/aaliyan-ahmed-rao