Thirupathi Radhakrishnan

Thirupathi Radhakrishnan

Principal AI Architect
UAE
Kannada, Tamil, English
منفتح على العروض
Full-time On-site

نبذة عني

Fifteen years across the ML lifecycle, from production risk and fraud models to enterprise MLOps and LLMOps platforms, now architecting an enterprise agentic AI platform for a large regulated organisation in the UAE. Des…

الخبرة

Strategic AI Architect

DEWA (Dubai Electricity and Water Authority) · Dubai
Mar 2026 - حتى الآن · 8 أشهر

<p>Architecting DEWA's enterprise agentic AI platform — the governed, reusable foundation on which the internal agent factory and business departments build and run agentic systems — working forward-deployed with those teams as the platform's customers.</p><p> Delivered Phase 1: harness engineering and the observability layer live, consumed by the agent factory and by departmental agentic systems running under harness control.</p><p> Completed the platform design — agent orchestration, tool and API gateway, memory and state, reusable harness patterns for multi-use-case deployment — now in development.</p><p> Designing the control plane: evaluation gating, AI governance, and token economics tied to value realisation.</p><p> Building the enterprise MLOps platform beneath it — CI/CD/CT, model registry, vLLM inference serving, drift and latency monitoring — standardising training, deployment and promotion across the AI portfolio.</p><p> Defining reference architecture and standards in alignment with the CAIO and platform leadership so delivery teams onboard onto a common foundation rather than build bespoke.</p>

Principal AI Architect

Xebia Middle East Software LLC · Dubai
May 2025 - Feb 2026 · 9 أشهر

<p>Led enterprise AI architecture for a regulated client, taking the MLOps and LLMOps platform from HLD/LLD through live operation across Dev, UAT and Production — embedded with the client's engineering and security teams from first design to go-live.</p><p> Configured vLLM model serving for production GenAI workloads alongside RAG pipelines and agentic orchestration (LangGraph, AutoGen), with evaluation, guardrails and versioning built in.</p><p> Engineered the model lifecycle toolchain: MLflow registry, automated training and validation, drift detection, retraining triggers and controlled promotion gates.</p><p> Defined platform standards and CoE frameworks that let multiple teams reuse pipelines and patterns; built the infrastructure as code with Terraform/Terragrunt on Azure and AWS, integrated with GitLab CI/CD and secure runner execution.</p><p> Delivered a production-ready enterprise MLOps and LLMOps platform end to end, enabling governed AI adoption across teams with repeatable, auditable deployments.</p>

Lead Data Scientist

Publicis Sapient · Bangalore
Apr 2024 - Dec 2024 · 7 أشهر

<p>Led end-to-end delivery of production ML and GenAI systems for enterprise clients, forward-deployed with risk, compliance and audit teams from problem framing through deployment, monitoring and automated retraining.</p><p> Built and operationalised MLOps and LLMOps pipelines — CI/CD, model and prompt versioning, real-time monitoring and rollback — for repeatable, auditable releases.</p><p> Engineered a cost-efficient, sustainability-aware architecture that minimised compute and storage overhead while holding enterprise-grade quality and traceability.</p><p> Delivered a GenAI-powered Intelligent Document Processing platform on LLMs and RAG (ingestion, classification, extraction, validation) that cut document retrieval time by ~92% at enterprise scale.</p>

Architect – Technology (AI & Data Platforms)

Virtusa · Bangalore
Jan 2023 - Apr 2023 · 3 أشهر

<p>Designed AI analytics and cloud-native data platforms for retail and enterprise clients on Azure, owning architecture and system-design decisions end to end.</p><p> Engineered large-scale retail POS data pipelines on Azure Databricks and built sales-forecasting models using Gaussian Process Regression, LSTM sequence models and Azure AutoML.</p><p> Improved sales-forecasting accuracy by ~15% and cut POS data processing time by ~40%.</p>

Lead Data Scientist – Risk & Fraud Analytics

Mirafra Technologies · Bangalore
Oct 2021 - Jan 2023 · 1 سنة 3 أشهر

<p>Directed a data science team delivering production machine learning for fraud detection and credit risk in a regulated lending environment, embedded with the client's risk and engineering teams from problem framing to live decisioning.</p><p> Took models from notebook to production — feature pipelines, rigorous evaluation, deployment support, post-launch monitoring — and established repeatable training workflows, versioned artefacts and validation gates with engineering.</p><p> Reduced fraudulent personal loan applications by 25% at ~93% model recall, tightening portfolio risk control and cutting manual review load.</p>

Founder & Data Science Specialist

Tempworks Services · Bangalore
Aug 2016 - Aug 2021 · 5 سنوات

<p>Ran a data science consultancy for five years delivering custom ML for risk, decisioning and profitability — forward-deployed inside each client's business from problem framing to models running in live, real-time decisioning systems.</p><p> Took every engagement past proof of concept into production, with integration into operational workflows and hand-over the client's teams could run.</p><p> Reduced credit losses by ~20% and improved profitability by ~18% through deployed credit-risk and borrower-behaviour models.</p>

Senior Account Manager

Randstad India · Bangalore
Jan 2014 - Mar 2015 · 1 سنة 1 شهر

<p>Managed strategic relationships with enterprise clients, driving growth through targeted account expansion.</p>

Branch Head – Staffing

Gi Group · Bengaluru , Karnataka , India
Jun 2012 - Jan 2014 · 1 سنة 7 أشهر

<p>Managed full-scope branch operations and staffing delivery, driving process optimisation and improved turnaround times.</p>

المهارات

Multi-agent orchestration LangGraph AutoGen CrewAI harness engineering tool/API gateways agent memory state management human-in-the-loop routing evaluation gating token economics LLM application architecture RAG pipelines vector databases embedding pipelines prompt engineering context engineering guardrails LLM evaluation benchmarking vLLM FastAPI gRPC batch inference GPU utilisation cost optimisation latency tuning throughput tuning MLflow Kubeflow Argo Workflows CI/CD CT GitLab CI/CD model promotion gates drift detection concept decay monitoring automated retraining Docker Kubernetes AKS EKS Terraform Terragrunt Azure AI Foundry Azure OpenAI AWS private endpoints network isolation
الإبلاغ عن هذا الملف الشخصي؟