Advisory
AI product & advisory work
Lead an AI advisory and build practice that partners with founders, CTOs, and product teams on GenAI and agentic adoption from strategy to production, with around 85% of engagements reaching production. The novelty comes from stitching together different components into a unique solution: multiple GenAI-powered products and agentic solutions built from 0 to 1 across Industrial, Fashion, Real Estate, ESG, Insurance, Retail, and Marketing, deep research on large datasets, and product AI copilots.
Sep 2024 to Mar 2026
SAM Corporate: AI Advisory & Solution Architect
Built an AI insurance early-fraud-detection agent from ideation to MVP and an ESG metrics-extraction pipeline that improved accuracy from 30% to 80%+, delivered as a full platform with human-in-the-loop validation, automated insights, and NL2SQL chat.
Dec 2025 to Feb 2026
Zubera: AI Strategy & Advisory
Coached a 25-person product, engineering, UX, and data team on the move from traditional to AI-native, delivering product features, AI-native journeys, prototypes, and a pitch deck while aligning data readiness to business outcomes.
Nov 2024 to Nov 2025
Lyzr AI: AI Advisory & Solution Architect
Led custom-agent design, use-case discovery, pipeline architecture, evaluation, and guardrails for enterprise clients (NTT, Accenture, Publicis, AirAsia, Willis Towers Watson) across Travel, E-commerce, HR, DevOps, Healthcare, and ITSM.
May 2024 to Jan 2026
Fitumi.ai: Strategic Technical Advisor & Vision Solution Architect
Conceptualized and shaped a fashion vision product that generates realistic on-model product imagery using an ensemble of LLM vision models, reducing sampling and product-development cost.
Apr 2024 to Jan 2025
Proplens AI: AI Strategist & Solution Architect
Built V1 of a real estate GenAI product for sales, marketing, and customer support, iterating on consistency, accuracy, and latency with Responsible AI adoption.
Ongoing
SB Solutions, AuditOne GmbH, ServCrust
Agentic retail systems and a Txt2SQL solution at 80% accuracy on complex multi-table schemas (SB Solutions); co-authored the STARED AI audit framework covering security, technical assessment, regulatory compliance, ethics, and data governance (AuditOne GmbH); a construction-procurement AI blueprint (ServCrust).
Have a similar engagement in mind? Let's scope it together.
Book a free call →Leadership & C-suite workshops
For corporates, enterprises, and SMEs deciding whether to build or buy
Workshops for leadership and C-suite teams on AI-native transformation, Claude Code and agentic coding, AI security, and responsible AI practices, built to sharpen the build-vs-buy call before budget is committed to a vendor or a build.
We won't charge for training alone. We travel through the transformation with your team to make sure it's practiced and followed, not just presented.
🤝 If this resonates, let's connect, I'll help you identify the one readiness gap costing your enterprise the most.
Current consulting offerings
Eight ways to bring Phygitalytics in
- 1.GenAI use case discovery & prioritisation
- 2.Enterprise GenAI adoption & cross-domain solution design
- 3.Agentic solution architecture, data integration, sales enablement
- 4.Code review, technical mentoring, alignment
- 5.Data readiness → AI capability → business outcomes → solutioning
- 6.AI security, Responsible AI, AI audit
- 7.ESG + AI intelligence platform
- 8.Enterprise AI upskilling & mentoring
Domains
Operating principles
How Sivaram thinks about AI adoption
CAP vs CAL
Like the CAP theorem’s Consistency, Availability, and Partition tolerance, GenAI systems face a similar trade-off in Consistency, Accuracy, and Latency (CAL).
AGI, redefined
AGI = Automated, Augmented, and Guided Intelligence, not Autonomous. Creativity and imagination remain human skills.
Readiness over awareness
✅ GenAI-First Leadership + Trained Team + Best Practices = Productivity. ✅ GenAI-Ready Leadership + Pilots + Trained Team = Stability. ❌ GenAI-Aware Leadership + Aggressive Adoption + AI-Aware Team = Chaos.
Why AI adoption really fails
AI doesn’t fail because of models. It fails because leaders lack AI literacy, developers lack domain literacy, and architects lack Responsible AI literacy.
AI audit, security & governance
The STARED AI audit framework, and input to the UN Global Dialogue on AI Governance
Co-authored AuditOne's STARED framework (Security, Technical Assessment, Regulatory Compliance, Ethics, Data Governance) for auditing AI at the function level, and Phygitalytics' submission was published among the written contributions to the UN Global Dialogue on AI Governance. Full AI audit, security, red-teaming, tooling, and governance detail lives on the dedicated Responsible AI page.
STARED audit, AI security, red-teaming, evaluation tooling, and governance, all in one place.
See the Responsible AI page →Field lessons
Working principles, with receipts
First principles thinking
Structure each client’s problem uniquely, question assumptions, and connect fundamentals to innovate under time pressure. For Proplens AI, this meant reimagining real-estate sales, marketing, and support workflows using Azure, Claude Sonnet 3.5, GPT-4, and Neo4j.
Hunger for impact
Reuse insight from past experiments to derive tangible products fast. For an industrial data-management project, prior data-catalog work let the team scale an LLM + custom-NER system across supply chain and procurement quickly.
Depth of analysis
Prototype and test reasonably. For Fitumi (vision + fashion AI), rapid prototyping across Stability AI, OpenAI, and Gemini identified the most effective way to cut sampling and marketing cost.
Brief, direct communication
Executives get synthesized insight with clear decision points. For AuditOne GmbH, recommendations on model monitoring, content moderation, and compliance stayed concise and example-backed.
Centering on strategy
Align AI initiatives to business objectives, not just current products. For Unilever’s meal-waste-reduction project, GenAI was framed around sustainability strategy, not just a feature.
Spotting false positives in GenAI implementation
- A demo that works may not be the solution you need. Validate it against real-world scenarios.
- Do not trust the LLM until your first 100 users are happy with it. User feedback is invaluable.
- Do not rely on benchmarks that don’t reflect your data. Trust your own metrics.
- Do not force-fit an LLM use case to get a promotion. Choose it only if it genuinely adds value.
- Prioritize accuracy first, then reduce costs. Achieving everything at once is unrealistic.
- If someone promises a working solution in one month, be cautious. It might be a prototype, not production-grade.
The AI consulting journey
Most AI consulting fails before the model even starts
The real problem is rarely the model. It's misalignment between what's asked, what's real, and what the team is ready for, the same thinking behind the Readiness over Awareness adoption model: listen deeply, expose the gap, assess talent, evaluate cost, align the solution, reset data thinking, and educate responsible AI before scaling.
Not sure which stage your team is at? A free 30-minute call will place it.
Book a free call →Curriculum & faculty work
Teaching ML and GenAI at scale
- Berkeley Haas, via Emeritus: AI and GenAI business strategy for executive education
- NTU Singapore: Data Science in Finance
- ISB: Leadership with AI, GenAI assignments, scenarios, and capstone projects
- IIM Kozhikode: Supply Chain and GenAI offerings and adoption recommendations
- upGrad: 178 sessions, about 350 hours of Deep Learning, Vision, NLP, and GenAI
Bring Phygitalytics into your product roadmap
From zero AI skills to shipped GenAI experiments, let's talk about your product.