Senior AI/ML Engineer

Description :

Unlock your dream career with us. Connect with renowned employers, enjoy top-notch facilities, and thrive in a vibrant community.

Position : Sr. AI/ML Engineer

Work Days : 5 days

Years of Experience : 6- 7 Year

Community Location : Pune / Noida

Employment Type : Full Time

Company Overview :

McKinley Rice is not just a company; it's a dynamic community, the next evolutionary step in professional development. Spiritually, we're a hub where individuals and companies converge to unleash their full potential. Organizationally, we are a conglomerate composed of various entities, each contributing to the larger narrative of global excellence.

Redrob by McKinley Rice: Redefining Prospecting in the Modern Sales Era

Backed by a $14 million Series A funding from leading Korean & US VCs, Redrob is building the next frontier in global outbound sales. Were not just another databasewere a platform designed to eliminate the chaos of traditional prospecting. In a world where sales leaders chase meetings and deals through outdated CRMs, fragmented tools, and costly lead-gen platforms, Redrob provides a unified solution that brings everything under one roof.

Inspired by the breakthroughs of Salesforce, LinkedIn, and HubSpot, were creating a future where anyone, not just enterprise giants, can access real-time, high-quality data on 700 M+ decision-makers, all in just a few clicks.

At Redrob, we believe the way businesses find and engage prospects is broken. Sales teams deserve better than recycled data, clunky workflows, and opaque credit-based systems. Thats why weve built a seamless engine for :

1. Precision prospecting

2. Data enrichment from 16+ premium sources

3. AI-driven workflows to book more meetings, faster

Were not just streamlining outbound were making it smarter, scalable, and accessible. Whether youre an ambitious startup or a scaled SaaS company, Redrob is your growth copilot for unlocking warm conversations with the right people, globally.

How to become a part of our community :

Step 1: 30-Minute Recruiter Screen (Virtual)

Step 2: Technical Interview (Virtual)

Step 3: Be a part of the team at McKinley Rice!

Experience :

Duties you'll be entrusted with :

- Architect and own end-to-end LLM-powered AI solutions for AI Connect from proof-of-concept through production at millions-of-users scale

- Design scalable inference serving infrastructure using vLLM, Ray Serve, BentoML, TGI, or equivalent with auto-scaling, load balancing, and failover

- Make build-vs-buy decisions on model selection (open-source vs proprietary LLMs) based on cost-per-query, latency SLAs, accuracy, and compliance trade-offs

- Define and standardise AI system architecture patterns RAG pipelines, agent orchestration, multi-model workflows, and model routing strategies

- Design multi-region serving architectures for low-latency global access

- Fine-tune and adapt open-source and proprietary LLMs (GPT-4, Claude, Llama, Mistral) for Redrobs prospecting and enrichment use cases

- Build and optimise RAG (Retrieval-Augmented Generation) pipelines for high-precision, low-latency responses over large datasets

- Implement advanced prompt engineering, chain-of-thought reasoning, and agent-based AI systems using LangChain, LlamaIndex, or custom frameworks

- Optimise model inference for cost and latency quantisation (GPTQ, AWQ, GGUF), speculative decoding, continuous batching, KV-cache optimisation, and efficient GPU utilisation

- Own the full model deployment pipeline from model registry to production endpoint, including blue-green deployments, canary rollouts, and zero-downtime model swaps

- Implement model versioning and artifact management (MLflow, Weights & Biases, or custom) so every production model is traceable, reproducible, and rollback-ready

- Build automated deployment gates eval benchmarks, latency checks, cost projections, and safety tests must pass before any model reaches production traffic

- Design and operate multi-model serving infrastructure where different use cases route to different models (e.g., fast/cheap for classification, powerful for generation)

- Manage model A/B testing and progressive rollout infrastructure to validate model changes on live traffic before full deployment.

- Own GPU fleet management instance selection, spot vs reserved strategy, GPU utilisation monitoring, and cost optimisation across the inference fleet

- Implement and tune adaptive batching, request queuing, and priority-based routing to maintain latency SLAs under variable load (burst traffic, peak hours)

- Build real-time inference observability per-model latency (P50/P95/P99), throughput, error rates, token usage, cost-per-query, and GPU memory/utilisation dashboards

- Implement inference caching strategies (semantic caching, prompt caching, KV-cache sharing) to reduce redundant computation and cost

- Design and enforce rate limiting, quota management, and graceful degradation strategies for AI endpoints serving millions of concurrent users

- Own the full LLM lifecycle training, evaluation, deployment, monitoring, drift detection, retraining triggers, deprecation, and decommissioning

- Build CI/CD pipelines for AI/ML application delivery, including automated evaluation suites, model versioning, and progressive rollout to production

- Implement output quality monitoring and drift detection for LLM systems hallucination detection, response quality scoring, and automated alerts for degradation

- Design and operate model rollback procedures when a new model underperforms, the system must recover to the previous version within minutes, not hours

- Build cost management infrastructure per-model, per-feature, and per-customer cost tracking with automated alerts and budget enforcement

- Own Infrastructure as Code (Terraform, Pulumi, or CloudFormation) for all AI/ML infrastructure GPU clusters, model serving endpoints, vector databases, and supporting services

- Manage Kubernetes (EKS) with GPU scheduling NVIDIA device plugin, time-slicing, MIG, node pools, and auto-scaling policies tuned for inference workloads

- Containerise and deploy AI workloads on AWS (ECS/EKS, SageMaker, GPU instances) with production-grade networking, security groups, and IAM policies

- Build and maintain auto-scaling policies that respond to inference-specific signals (request queue depth, GPU utilisation, batch queue size) not just CPU/memory

- Own incident response for AI systems be capable of diagnosing and resolving GPU OOM errors, inference latency spikes, model serving crashes, and deployment failures

- Implement disaster recovery and high-availability patterns for AI infrastructure multi-AZ deployments, health checks, circuit breakers, and automated failover

- Expose AI capabilities as well-documented APIs and MCP (Model Context Protocol) tools that integrate with existing backend services

- Collaborate closely with backend leads, data engineering, and frontend teams to ship AI features end-to end

- Work with the cybersecurity consultant to ensure AI systems meet security and data privacy requirements

- Mentor junior AI engineers (5 on the team) conduct code reviews, drive design discussions, and raise the teams technical bar

- Stay current with the rapidly evolving LLM landscape and evaluate new models, techniques, and tools for production applicability

- Drive proof-of-concepts for emerging capabilities multi-agent orchestration, multimodal AI, agentic workflows

- Maintain clear documentation, experiment logs, and knowledge-sharing practices within the AI team

Expectations from you :

Mandatory Requirements :

- 6- 7 years of professional experience in ML/AI engineering, with at least 2+ years focused on LLM-based applications in production

- Large-scale deployment : Must have led or been a core contributor to at least one AI/ML system serving production traffic at significant scale (100K+ requests per minute or equivalent throughput)

- Deep hands-on experience with LLM frameworks Hugging Face Transformers, LangChain, LlamaIndex.

- Experience fine-tuning both open-source (Llama, Mistral) and proprietary models (GPT, Claude)

- Proven experience deploying LLM models to production model versioning, blue-green/canary deployments, rollback procedures, and A/B testing on live inference endpoints

- Hands-on with GPU fleet management, adaptive batching, request queuing, inference caching, and cost per-query optimisation at scale

- Production experience building RAG pipelines and agent-based systems, including retrieval strategy design, chunking, embedding models, and vector databases (Pinecone, Weaviate, Qdrant, pgvector)

- Hands-on with MLflow, Weights & Biases, Kubeflow, or equivalent for experiment tracking, model registry, deployment automation, and lifecycle management

- Proficient with Terraform/Pulumi/CloudFormation for infrastructure provisioning. Comfortable managing Kubernetes clusters with GPU workloads, writing Helm charts, and debugging infrastructure issues

- Strong experience deploying ML workloads on AWS (preferred) or GCP Docker, Kubernetes (EKS/GKE), GPU instance management, auto-scaling, and cost optimisation

- Strong proficiency in Python. Working knowledge of TypeScript/JavaScript for API integration. Bash scripting for automation

- Advanced prompt design, chain-of-thought, few-shot learning, and systematic prompt evaluation techniques

- Experience setting up monitoring for AI systems latency dashboards, output quality tracking, cost monitoring, and alerting (Prometheus, Grafana, Datadog, or equivalent)

- Ability to articulate complex AI concepts to non-ML stakeholders, write clear design docs and runbooks, and mentor effectively

- Bachelors degree or higher in Computer Science, AI/ML, Data Science, Mathematics, or related fields

Optional Skills :

- Experience with the Model Context Protocol (MCP) and exposing AI capabilities as MCP tools/resources.

- Familiarity with NestJS or similar backend frameworks for AI-backend integration

- Experience with causal inference models (SCM, counterfactual analysis) or Explainable AI (XAI)

- Hands-on with multi-agent orchestration frameworks and multimodal AI integrations

- Experience with GPU scheduling in Kubernetes (NVIDIA device plugin, time-slicing, MIG)

- Contributions to open-source AI/ML, MCP, or LLM tooling projects

- Experience with data enrichment, entity resolution, or knowledge graph construction at scale

- Familiarity with sales-tech or B2B data products

- Experience with multi-region deployment and global traffic management for AI services

- Hands-on with FinOps practices for AI/ML workloads reserved instances, spot fleet management, cost allocation tagging