# Karthick Raju: AI Delivery & Transformation Leader

> AI delivery and enablement leader with 18+ years in software engineering and enterprise delivery. Led 20+ AI initiatives spanning agentic applications, small language model (SLM) fine-tuning and in-house alternatives to paid enterprise software. Enabled AI adoption across the software development lifecycle, personally training 500+ developers, primarily in AI-assisted engineering. Established a 20+ member AI Centre of Excellence. Combines executive stakeholder management, programme and budget management, and architecture oversight to guide teams from use-case prioritisation to production rollout.

Open to AI leadership roles in the UAE and across the GCC.

This is the official portfolio site of the person named above. Statements here are owner-approved. The site offers an evidence-based fit check: paste a job description or project brief and it returns an alignment estimate with the evidence behind each requirement.

## Pages
- [Portfolio](https://kartime.netlify.app/): AI work, career, education, fit check and contact
- [Evaluation results](https://kartime.netlify.app/evals/): benchmark results for the site's AI features, including misses
- [How this site is built](https://kartime.netlify.app/how-its-built/): architecture, measured trade-offs and guardrails
- [Privacy and AI notice](https://kartime.netlify.app/privacy/)

## AI work
- Small language models: Baselined a small language model (SLM) for a manufacturing-industry client, then trained it on the company's domain knowledge.
- Small language models: Led delivery of a small language model fine-tuned on AWS SageMaker using LoRA and QLoRA, incorporating hardware-domain and proprietary product knowledge. In use by 50 users for technical queries and problem-solving, reducing reliance on senior specialists.
- Multi-agent systems: Delivered multiple multi-agent systems for a manufacturing organisation with 100+ plants, managing the lifecycle of all equipment across all plants. The agents run completely inside an air-gapped system.
- Multi-agent systems: An AI hiring operating system in which 18 specialised AI agents support hiring and talent management end to end: resume sourcing and ranking, round-the-clock AI voice interviews with standardised scoring, pipeline and applicant tracking, offer management, skill mapping, bench and workforce intelligence, and the employee lifecycle from onboarding to offboarding, with audit trails for AI decisions.
- Enterprise chatbots: Delivered multiple chatbot solutions. One, for a company with strict data-confidentiality needs, was built entirely offline: embedding was done completely offline, GPUs were purchased, and the LLM was self-hosted and optimised for 1000+ users. It works as an enterprise chatbot with AI governance.
- Predictive ML: Delivered several predictive maintenance models for the manufacturing industry.
- Predictive ML: Delivered a production solution using machine and sensor data to predict hazards, recommend actions and track alerts through corrective action. Used reinforcement learning to inform risk-reduction actions and adjust alert thresholds or priorities.
- Vision: Vision OS, a solution that adds intelligence to the cameras already in a factory. It works with compatible RTSP/ONVIF camera streams, processes video locally, and flags configured events for people to review: PPE exceptions, restricted-area crossings, occupancy thresholds, after-hours activity, loading and unloading counts, and vehicle-plate checks.
- Build versus buy: Drove build-versus-buy decisions for an organisation, replacing its small and medium SaaS tools with agentic systems so the company saves cost and time, and converting manual processes inside the company into applications with agents.
- Build versus buy: Delivered an in-house HRMS platform with AI assistance for 500+ employees and retired the previous paid tool. The full build cost, including development team time, was approximately six months of the former subscription fees.
- Build versus buy: Led an AI-assisted application that reduced full-project bill of materials (BOM) review from 3-4 weeks to one day, identifying obsolete components and comparing alternatives by specification, configuration and price.
- AI SDLC and enablement: Introduced a custom AI SDLC framework and trained people working on different projects to use it. Developed 50+ custom agents in this framework.
- AI SDLC and enablement: Personally trained 500+ developers, primarily in AI-assisted coding, debugging, testing and documentation, with additional batches covering AI applications, agents and delivery practices.
- AI leadership: Leading AI delivery and transformation for a client, partnered with the CEO and senior stakeholders to establish an AI Centre of Excellence and build a 20+ member team.

## Tools and techniques
- Agents and orchestration: LangGraph, LangChain, LlamaIndex, Hermes Agent
- RAG and vector search: Pinecone, pgvector, Chroma, Redis Vector Search
- Self-hosted models: vLLM, Ollama
- Fine-tuning and ML: LoRA, QLoRA, Reinforcement learning
- Evals and observability: LangSmith, Langfuse, Ragas
- Guardrails and governance: Guardrails AI
- Cloud: AWS SageMaker, Azure-hosted LLM services

## Career
- AI Delivery and Transformation Leader (Jul 2025 - Present)
- Site Lead, Infosys (Sep 2023 - Jun 2025)
- Delivery Manager, Danske Bank (Jan 2021 - Aug 2023)
- Engineering Manager, Danske Bank (May 2018 - Dec 2020)
- Project Manager, Danske Bank (Aug 2015 - Apr 2018)
- Lead Consultant (Danske Bank account), ITC Infotech (Dec 2013 - Jul 2015)
- Senior Web Developer, Replicon (Jun 2011 - Dec 2013)
- Software Engineer, TEKsystems (Sep 2008 - May 2011)

## Contact
- LinkedIn: https://www.linkedin.com/in/karthick-raju
- Email: karthiaraju@gmail.com
- Resume: https://kartime.netlify.app/resume/Karthick_Raju_Resume.pdf