Senior Software Engineer

Key Responsibilities

Architecture & System Design

• Architect and deploy end-to-end AI systems — from data pipelines to model serving.
• Design modular SDKs for multi-provider AI integration (OpenAI, Claude, Gemini, LLaMA).
• Lead decision-making on cloud vs self-hosted LLM deployment (Ollama, vLLM, TGI).
• Guide infrastructure design for scalability, observability, and cost efficiency using GPU clusters, Ray, or KServe.
• Collaborate with backend, MLOps, and infra teams to ensure high availability and low latency across AI workloads.

Core ML / DL Development

• Train and fine-tune models (CNN, RNN, Transformers) across text, vision, and speech domains.
• Implement LoRA / PEFT fine-tuning for custom LLMs, embedding models, and instruction-tuned variants.
• Work with open-source and proprietary model repositories (Hugging Face, Kaggle, Hugging Face Spaces).
• Optimize model architectures for inference performance, quantization, and memory efficiency.
• Conduct A/B testing, cross-validation, and human evaluation on model outputs.
• Build internal evaluation benchmarks and dataset management pipelines for consistent model scoring and comparison.

Data & Dataset Engineering

• Curate, clean, and version-control datasets for text, image, and audio modalities.
• Build pipelines for data labelling, augmentation, and validation using Airflow / Prefect.
• Create and manage feature stores, embedding repositories, and dataset registries.
• Leverage open datasets (e.g., Common Crawl, LAION, OpenImages, LibriSpeech) and integrate custom enterprise datasets.
• Ensure data governance, bias checks, and PII anonymization using Presidio or custom filters.

AI Ops & Deployment

• Automate model workflows with MLflow, Kubeflow, or Vertex AI for experiment tracking and versioning.
• Lead model deployment with vLLM, TGI, or TorchServe, ensuring optimized GPU/TPU utilization.
• Set up continuous evaluation pipelines for model drift, bias, and quality decay using EvidentlyAI and Prometheus.
• Leverage open datasets (e.g., Common Crawl, LAION, OpenImages, LibriSpeech) and integrate custom enterprise datasets.
• Drive adoption of model registries and model cards for transparency and reproducibility.

Team & Technical Leadership

• Mentor and review the work of AI/ML Engineers I & II.
• Collaborate with product, design, and research teams to translate business needs into AI roadmaps.
• Lead POCs and experiments for emerging AI verticals (e.g., multimodal, video, robotics, IoT intelligence).
• Present internal demos, AI reports, and architectural documentation to leadership and clients

Core Skills Required

• Programming: Expert-level Python, with a deep understanding of OOP, async, and design patterns
• Frameworks: PyTorch, TensorFlow, Hugging Face Transformers, LangChain,LlamaIndex.
• Model Ops: MLflow, KServe, TorchServe, vLLM, TGI.
• Data Stack: Airflow / Prefect, pgvector, Milvus, Pinecone, FOSS, PostgreSQL.
• Infra: Docker, Kubernetes, Ray, GPU servers, Cloud AI (Vertex AI, Bedrock, Azure).
• Evaluation & Metrics: Familiarity with BLEU, ROUGE, and latency/throughput metrics for AI models.
• Security: Secure Vaults, Microsoft Presidio, Fairlearn / AIF360 awareness for data and bias governance.

Good-to-Have Skills

• Experience with distributed training, quantization, and mixed-precision optimization.
• Experience with model compression, distillation, or low-rank adaptation for efficiency.
• Contribution to open-source AI frameworks or Hugging Face Spaces.
• Research exposure in LLM alignment, prompt optimization, or multimodal reasoning.
  • Understanding of AI cost governance, observability, and MLOps automation.