AI · DATA · SYSTEMS

Helping enterprises build AI and data systems that earn their keep.

I'm Nishant — a Data & AI consultant with fifteen years inside large enterprises. I help leaders turn ambition into systems that actually ship:

Shaping Data & AI strategy, architecture and roadmaps
Leading pre-sales, discovery and solutioning for enterprise programs
Designing GenAI, agentic and analytics platforms that survive production
Writing down what I learn so the next engagement starts smarter

Read the journal Wander the ideas

Nishant Sehgal

WRITING·ADVISING·STILL LEARNING

IDEAS I KEEP RETURNING TO

Six threads I can't seem to put down.

Not a service menu — the questions I find myself writing, sketching and arguing about with whoever will listen.

Data & AI strategy

Data & AI roadmaps tied to measurable business outcomes
Cloud modernisation strategies across Azure, AWS and GCP
Governance frameworks that survive a budget cycle

Solution architecture

End-to-end Lakehouse blueprints on Databricks, Fabric and Snowflake
Reference architectures for BI, ML and GenAI workloads
Patterns that scale from pilot to enterprise rollout

Pre-sales & advisory

Discovery workshops and architecture reviews
Effort estimation, RFP/RFI responses and SOWs
Translating business intent into deliverable scope

GenAI & agentic systems

RAG, copilots and agentic workflows for the enterprise
Evaluation, guardrails and responsible AI controls
Knowledge management grounded in real source-of-truth data

Platform modernisation

Lakehouse, warehouse and self-service analytics platforms
Data contracts, lineage and catalog-driven governance
Migration from legacy ETL and BI to cloud-native stacks

Decision intelligence

Executive dashboards and KPI frameworks leaders trust
Embedded analytics and API-driven reporting
Closing the loop from insight to operational action

ABOUT ME

A practitioner with a notebook habit.

I work with leaders and teams on the harder edges of Data & AI — strategy, architecture, pre-sales, and the patient work of getting systems into production.

My career has moved across every layer of the enterprise data stack: from data warehousing, ETL and BI in the early years, through cloud-scale data platforms and machine learning, and into today's generative AI, copilots and agentic systems.

Most weeks I'm inside engagements with large enterprises — running discovery, shaping target architectures, owning the technical pre-sales, and staying close to delivery so the build matches the promise. The rest of the time, I'm here, turning patterns I see across industries into essays for fellow practitioners and decision-makers.

I work with a deliberately small number of clients across retail, manufacturing, financial services, automotive, consumer goods and telecom. Fewer engagements, deeper context, more honest advice.

Nishant Sehgal

15+

Years in Data & AI

Industry verticals

20+

Enterprise programs

Hyperscaler alliances

WHAT I VALUE

Curiosity. Rigour. Honest writing over polished noise.

WORLDS I'VE WORKED IN

Eight industries, one stubborn curiosity.

Each domain teaches a different lesson about how AI behaves in the wild. Here's what I've taken from each.

Retail

Merchandising · personalisation · loyalty

Demand forecasting and assortment optimisation
Personalisation and loyalty analytics at scale
Unified customer view across stores and digital

Manufacturing

OT/IT · predictive ops · quality

OT/IT convergence for connected operations
Predictive maintenance and asset performance
Quality analytics across global production lines

Automotive

Connected vehicle · aftersales · dealer analytics

Connected-vehicle telemetry and data platforms
Aftersales intelligence and warranty analytics
Dealer performance and sales operations reporting

Consumer Goods

Demand · supply · trade promotion

Demand sensing and S&OP analytics
End-to-end supply-chain visibility
Trade-promotion effectiveness and revenue growth

Financial Services

Risk · regulatory · advisory copilots

Risk, finance and regulatory data platforms
Customer 360 for retail and corporate banking
GenAI copilots for advisors and operations teams

E-Commerce

Catalog · search · clickstream

Catalog enrichment and search relevance
Clickstream pipelines for real-time decisioning
Experimentation platforms and A/B at scale

Logistics

Network optimisation · visibility

Real-time shipment and fleet visibility
Route, load and network optimisation
Operational KPIs grounded in clean event data

Telecommunications

Churn · network · customer 360

Customer 360 and lifetime-value analytics
Churn prediction and next-best-action
Network performance and capacity analytics

THE TOOLCHAIN

Tools I've lived with — opinions included.

Stylised marks, real history. Each one is a system I've designed with, broken, and learned a few things from.

FOUNDATION MODELS & GENAI

OpenAI

Anthropic

Gemini

Microsoft Copilot

Mistral

Hugging Face

LAKEHOUSE & DATA PLATFORMS

Databricks

Microsoft Fabric

Snowflake

BigQuery

dbt

BI & ANALYTICS

Power BI

Tableau

Looker

Qlik

HYPERSCALERS

Microsoft Azure

AWS

Google Cloud

Databricks

AI ENGINEERING

LangChain

Pinecone

Vertex AI

Kafka

MLOPS & DELIVERY

Hugging Face

MLflow

Airflow

Weights & Biases

STRATEGIC ALLIANCES

Microsoft Azure

Databricks

AWS

Snowflake

Microsoft Fabric

Google Cloud

HOW I THINK ABOUT SYSTEMS

Six principles that keep earning their keep.

Not a methodology. A set of stubbornly useful ideas I find myself reaching for whenever a new AI system has to survive contact with an organisation.

Think in layers, ship in slices

Separate data, semantics, AI and experience layers
Each layer evolves on its own release cycle
Reduces re-platforming risk across multi-year programs

"Composability beats cleverness."

Data as a product

Clear owners, contracts, SLAs and lineage per dataset
Treat consumers as customers, not ticket queues
Stable foundations make AI on top feel less fragile

"Garbage in, garbage at scale."

Governance as guardrails, not gates

Policy, access and classification live close to the data
Catalog and lineage automated, not maintained by hand
Enables self-service without losing audit confidence

"Trust scales when policy is code."

Agentic AI as workflow

Design agents as orchestrated, observable workflows
Constrain tool use, memory and decision scope by intent
Measure outcomes with evals, not vibes

"Constrain the agent, free the user."

Pre-sales is a design discipline

Shape architecture and scope during discovery, not after SOW
Estimation grounded in patterns, not optimism
Align business case, delivery plan and technical reality

"Clarity early, change later — never the reverse."

Accelerators over heroics

Reusable reference architectures and code accelerators
Delivery frameworks that compress time-to-first-value
Turn each engagement into capability for the next

"Build the kit, then build the system."

THE JOURNEY SO FAR

Fifteen years of consulting across AI and data.

Each chapter was really about one lesson I needed to learn before the next one made sense.

2008 — 2014

Foundations in data

Cut my teeth on the plumbing layer of enterprise analytics — the unglamorous work that decides whether dashboards earn trust.

HIGHLIGHTS

Enterprise data warehousing on Teradata, Oracle and SQL Server
ETL pipelines, data modelling and operational reporting
First exposure to large-scale BI rollouts across business units
Learned that data quality and stewardship decide every outcome downstream

2015 — 2019

Into solution design

Moved from build to design — closer to the business, the workshops and the trade-offs that shape every program.

HIGHLIGHTS

Solution architecture for BI modernisation and cloud migrations
Programs across retail, manufacturing and financial services
Discovery workshops, requirement shaping and value framing
Began writing proposals and SOWs that mapped cleanly to delivery

2020 — 2023

Pre-sales, up close

Years inside enterprise pre-sales for cloud analytics and data modernisation — half engineering, half translation.

HIGHLIGHTS

Architecture reviews and effort estimation for large programs
RFP/RFI responses, SOWs and commercial constructs
Cross-platform work: Azure, AWS, GCP, Databricks, Snowflake
BI modernisation on Power BI, Tableau and embedded analytics

2024 — 2025

From projects to a practice

Helped shape an end-to-end Data & AI practice — projects became a repeatable capability with accelerators and alliances.

HIGHLIGHTS

Practice design across data engineering, BI, ML and GenAI
Agentic systems and enterprise knowledge management offerings
Reference architectures, accelerators and delivery playbooks
Alliances with Microsoft, Databricks, AWS, Snowflake and Google

2026 — NOW

Writing it down

Still inside the work — advising leaders and teams while turning patterns from across engagements into essays and notes.

HIGHLIGHTS

Advisory on Data & AI strategy and platform modernisation
Cautious, evaluation-led rollout of agentic AI in production
Writing essays and field notes for practitioners and leaders
Keeping a small client list to keep the thinking honest

"If any of this resonates, I'd be glad to discuss the work and the thinking behind it."

Start a conversation