Chandra Tungathurthi
AI Consultant & Technical Architect · Sovereign Enterprise AI
I turn complex, traditionally expensive business problems into streamlined AI solutions: generative AI, multimodal data and LLMs that run where the client controls them. GDPR by architecture, not by policy.
Core competencies
- Sovereign enterprise AI: on-premise and EU-hosted LLMs, RAG, agentic AI
- Product and master data AI: PIM, DAM, PLM, MDM, ETIM/eCl@ss, classification, enrichment, translation at scale
- Large-scale LLM processing and cost control: batching, deduplication, per-record audit, budget caps
- LLM fine-tuning and on-device inference: quantisation, ONNX, WebGPU, models that run in the browser
- Pilot to production: Kubernetes, CI/CD, integration architecture, so AI systems run unattended
Project history
Mid-market manufacturers and distributors (via a German IT consultancy)
Lead AI Architect and AI consultant · Sovereign product-data AI: platforms and paid pilots
Built the AI practice and its platforms (large-scale LLM processing, agentic chatbots, enterprise AI search); sold and delivered paid pilots: translation of 200,000+ items, ETIM/eCl@ss classification, attribute completion, hero-image selection; on-premise or Open Telekom Cloud.
US health-tech startup (name on request)
AI Architect, prime technical consultant (close to founding engineer) · Agentic backend of an autonomous AI health agent, now released on iOS and Android
Conceptualised and built the LLM-orchestrated multi-agent backend: agents interpret patient queries, plan, connect to and fetch records, biomarkers, wearable and medication data, and draft interventions under physician-authored protocols. HIPAA-grade data handling.
Department store chain and medical technology manufacturer (via a Swiss product-data integration specialist)
Solutions Architect · PIM/ERP integration into Syndigo and AI-assisted attribute mapping
Integration solutions on the employer's integration platform; designed the UX of an AI-assisted attribute-mapping tool and took it to production for the department store chain, including infrastructure and release.
Confidential client
Technical Architect AI · RAG platform and enterprise AI search
RAG solution combining commercial and open-source LLMs; connector-based enterprise AI search; vision models for image-to-video.
Medical and scientific publisher (via a German digital consultancy)
Lead Applied AI Architect · AI-as-an-API platform for product data in Stibo STEP
After conventional AI attempts had failed, delivered classification and enrichment through the API-based AI platform built from zero to production; the client presented the results at the Stibo conference.
Agricultural trading group (via a German digital consultancy)
Integration Architect · Company-wide integration architecture
Re-platformed a hand-written microservice landscape onto a MuleSoft-based integration architecture on-premise; CI/CD for SAP Commerce with GitLab and GitHub.
Swiss food wholesale group (via a German digital consultancy)
Infrastructure / DevOps Architect · Overhaul of the group's e-commerce systems
Designed and built the entire Azure infrastructure and CI/CD deployment for one of the company's largest projects.
German mail-order retailer (via a German digital consultancy)
Infrastructure lead · On-premise infrastructure consolidation and first AI use case
Consolidated and cleaned up the on-premise landscape; introduced AI-based hero-image identification.
Skills
- AI systems
- Agentic systems and multi-agent LLM orchestration · RAG and enterprise AI search · LLM evaluation, prompt optimisation and human-in-the-loop · LLM fine-tuning, quantisation, ONNX export and on-device inference (WebGPU) · sovereign, self-hosted and EU-hosted open-weight models · multimodal (vision, speech) · large-scale batch LLM processing (idempotent, crash-safe, cost-optimised) · guardrails and cost control
- Product & master data
- PIM, DAM, PLM, MDM · Stibo STEP, Syndigo, SAP Commerce · ETIM / eCl@ss classification · enrichment, translation, hero-image selection, marketing text
- Platforms & infrastructure
- Kubernetes (EKS, AKS) · Azure, AWS, Open Telekom Cloud, on-premise · Terraform and Infrastructure as Code · CI/CD with Dagger, GitLab, GitHub · Prometheus, New Relic · MuleSoft integration architecture
Platforms designed & built
- Large-scale LLM processing platform: Millions of records through LLMs (architected and built in employment)
- CAIEP: Composable Agentic Integration & Execution Platform (own product)
- DocVault: Agentic Document Intelligence (own product)
- Offline RAG chatbot: Knowledge on demand, no internet, no per-prompt cost
- On-device AI twin: Fine-tuned bilingual LLM running entirely in the browser (own work)
How I work
Frame · Pilot · Evaluate · Productionize. Pilots in weeks, not quarters, with an honest go/no-go before serious budget is committed.
Languages
- English: native / bilingual
- German: working proficiency