HSM Edifice Construction Services
Senior AI/ML Engineer/ Manager
- Compensation
- ₹1.2M–₹1.5M / yr
- Location
- Nagpur
- Work setup
- Onsite
- Seniority
- Senior
- Employment
- Not specified
- Experience
- 2+ years
- Category
- Ml / AI
- Posted
Skills
Must have
- 5+ years of professional experience in machine learning, data science, or AI engineering, including 2+ years in an architect, lead, or senior IC role driving system design.
- Bachelor's or Master's in Computer Science, Engineering, Mathematics, or equivalent practical experience.
- Strong Python; production experience with PyTorch and/or TensorFlow.
- Demonstrated experience shipping generative AI systems to production RAG, agents, or fine-tuned LLMs not just prototypes or notebooks.
- Solid computer vision background: CNNs, transformer-based vision models, detection/segmentation architectures, and the practical realities of training on imperfect real-world data.
- Deep, hands-on AWS experience for ML workloads, including SageMaker and at least one of Bedrock, EKS, or Lambda-based inference.
- Experience with MLOps tooling: Docker, Kubernetes, CI/CD, model versioning (MLflow, Weights & Biases, or similar), and Infrastructure as Code (Terraform or CloudFormation).
- Proven ability to design for scale, latency, and cost you can reason about GPU utilization, throughput, and unit economics, not just accuracy metrics.
- Clear written and verbal communication; comfortable defending an architecture in front of a room.
Nice to have
- Design end-to-end AI/ML architectures covering data ingestion, feature stores, model training, inference serving, monitoring, and feedback loops.
- Produce architecture decision records (ADRs), reference designs, and technical roadmaps; drive design reviews with engineering and security stakeholders.
- Evaluate and select models, frameworks, and vendors; own build-vs-buy and open-source-vs-managed-service decisions.
- AWS certification (Machine Learning Specialty, or Solutions Architect Professional).
- Experience with vector databases (OpenSearch, Pinecone, Weaviate, pgvector, FAISS).
- Experience with LLM frameworks (LangChain, LlamaIndex, DSPy) and serving stacks (vLLM, TGI, Triton).
- Edge or real-time CV deployment (NVIDIA Jetson, GStreamer, RTSP pipelines).
- Distributed training experience (DeepSpeed, FSDP, Ray).
- Streaming and big-data experience (Kafka, Spark).
- Familiarity with responsible AI practice: bias evaluation, model cards, explainability, and emerging regulation (EU AI Act, NIST AI RMF).
- Domain exposure in [healthcare / manufacturing / retail / fintech / logistics].
- Open-source contributions, patents, or published research
Full description
Job descriptionRole & responsibilities:About the Role
We are looking for an AI Architect to own the technical design and delivery of our AI platform. This is a hands-on architecture role: you will translate business problems into production AI systems spanning generative AI, computer vision, and classical ML, and you will define how those systems are built, deployed, and scaled on AWS.
You will work across product, data engineering, and MLOps teams setting architectural standards, making build-vs-buy calls, and making sure what we ship is accurate, cost-efficient, secure, and maintainable.
Preferred candidate profile:Key ResponsibilitiesArchitecture & Design
- Design end-to-end AI/ML architectures covering data ingestion, feature stores, model training, inference serving, monitoring, and feedback loops.
- Produce architecture decision records (ADRs), reference designs, and technical roadmaps; drive design reviews with engineering and security stakeholders.
- Evaluate and select models, frameworks, and vendors; own build-vs-buy and open-source-vs-managed-service decisions.
Generative AI
- Architect LLM-based solutions: RAG pipelines, agentic workflows, tool/function calling, and multi-step orchestration.
- Design vector search and retrieval layers chunking strategy, embedding model selection, hybrid search, re-ranking.
- Implement fine-tuning, LoRA/PEFT, and prompt-optimization strategies where they beat prompting alone.
- Define evaluation frameworks for LLM output quality, hallucination rate, latency, and cost per request.
- Build in guardrails: prompt-injection defenses, PII redaction, content filtering, and human-in-the-loop review.
-
Computer Vision
- Design and deploy CV systems for [object detection / OCR & document extraction / image classification / segmentation / video analytics / quality inspection].
- Own the data pipeline for vision: annotation workflows, augmentation, class imbalance handling, and dataset versioning.
- Optimize models for the target environment quantization, pruning, distillation, ONNX/TensorRT conversion, and edge deployment where applicable.
- Work with multimodal models (vision-language) where they simplify or outperform traditional CV pipelines.
AWS & Platform
- Architect AI workloads on AWS using services such as SageMaker, Bedrock, Lambda, ECS/EKS, S3, Step Functions, API Gateway, Kinesis, and OpenSearch.
- Own cost architecture: instance/accelerator selection, spot strategy, autoscaling, caching, and token-spend controls.
- Implement CI/CD for models (MLOps): automated retraining, model registry, canary and blue-green rollouts, drift detection, and rollback.