Consulting
I am an Associate Professor of Marketing at IIM Calcutta, and I take on a small number of consulting engagements in AI and analytics. My clients so far include a government ministry, manufacturers, a retail chain, software and publishing-services firms, and early-stage ventures.
The work usually starts with a process rather than a tool. I map how a task is done today, decide which parts are repeatable and low-risk, and keep a person at the points where judgement matters. Then I build something small enough to measure, hand it over, and say plainly what it did and did not do. On this page some clients are named with their permission and others are described by sector. Where I have a measured result I give it. Where I do not, I leave the outcome blank rather than guess. Several engagements are still running, so I describe the method and leave out anything a client has not released.
I start with how the work is done today, then decide what to automate, what to assist and where a person stays in charge.
Where a task is narrow or data must stay in-house, I advise on and build small models trained for that one job.
I turn rules, documents and public registries into assistants that answer from a known source and withhold what they cannot support.
Agents that read operational data and produce a briefing, a list or a plan, with a person at the decisions that matter.
I test what works on the client's own data and report the failures next to the successes.
This is my original field. I advise on search, social, content and websites, and on how AI changes each of them.
Selected engagements
Some clients are named with their agreement. Others appear by sector only, because the work is live or commercially sensitive. Each one has a page of its own with the problem, the method, the numbers and their limits. Where no outcome is recorded, none was measured, and I would rather say so than imply one. The picture on each card illustrates the domain the work sits in; it is not a photograph of the client.
Government, urban development, tourism
A mixed-method impact assessment for the Ministry, asking whether Smart Cities Mission upgrades changed what visitors experience, and how many of them come.
94 smart-city offices surveyed, 1,075 tourists surveyed across three cities, 22 in-depth interviews. Signed off on 8 December 2025 as having fully met its objectives.
Read the case →
AI infrastructure advisory, data centres, events
Strategic advisor on marketing and AI strategy to a UAE advisory firm, and owner of the work that designed and built its public site and launched its flagship summit.
Live in production with 44 routes, including four published articles. No traffic, lead or registration figures.
Read the case →
Scholarly publishing production services
Twelve ranked automation opportunities, one shared architecture, and working prototypes: a reference and retraction checker, a books project-management system, and a rules-first manuscript-to-XML pipeline.
On a 16-item public gold set the reference checker resolved 13 of 13 DOIs and caught 5 of 5 retracted papers with a Retraction Watch mirror. A Qwen3-1.7B triage fine-tune scored 88.3% against 63.8% for a prompt baseline on 94 held-out queries.
Read the case →
Engineering design services, heating, cooling and electrical design for office buildings
Four hours per floor go into cleaning an architect's drawing by hand before engineering can start. I built and measured a prototype that does the tractable parts of that, and reported the part that is not solved.
Tested on 14 of the client's floors: areas and occupants matched for 100% of rooms, room type from a room name 95% correct, chairs located 93% of the time. Room boundaries from raw walls, 23.5%, which is not solved.
Read the case →
Contract manufacturing, aluminium castings and machined components
A quotation-rules knowledge pack for a custom GPT that reads a casting drawing, tabulates every dimension and tolerance, checks feasibility and writes the queries back to the customer.
No outcome was formally measured.
Read the case →
Retail, small-town and rural grocery
One command runs the chain from data. Roughly 50 agents read ERP exports and store WhatsApp chats on a local machine and return a dashboard, an Excel file and a short summary, with every number tagged to its source.
No outcome was formally measured.
Read the case →
Education, online courses, public sector
A data audit, learner segmentation and interactive dashboards for a government-backed learning platform, asking where learners and value move between the platform and its course partners.
No outcome was formally measured.
Read the case →
Enterprise software, financial risk and controls assurance
Positioning, messaging, information architecture and page-by-page copy for a risk and controls software firm, with a working prototype of the new site deployed to a protected preview.
No outcome was formally measured.
Read the case →
Financial services technology
Where small language models fit in loan origination software whose customers require that data stays in their own environment, and why the first thing to build is the measuring instrument, not the model.
No outcome was formally measured.
Read the case →
Healthcare and life sciences, brand and marketing
A generative AI introduction for a brand and marketing team in June 2023, built around what the technology could and could not do for their work, rather than around the tools of the month.
No outcome was formally measured.
Read the case →
Retail, pet supplies
A phased digital and e-commerce plan for a chain of physical pet stores, from website and search in phase one to a care app and recommendations later, plus the shape of the team needed to run it.
No outcome was formally measured.
Read the case →
Industrial technology, B2B
Search and LinkedIn demand generation for a firm selling machine and battery monitoring to factories: keyword families with matching landing pages, and a weekly content routine the team could actually sustain.
No outcome was formally measured.
Read the case →
Industrial manufacturing, with several group businesses
Branding for both customers and employees at a multi-business industrial group, with a LinkedIn competitor mapping against European peers and a review of the agency scope for a new site.
No outcome was formally measured.
Read the case →
Social enterprise, microfinance and livelihoods, rural India
The marketing module of a proposed mini-MBA for self-help group federation leaders, including using ChatGPT and Gemini in Hindi, by voice, as a business assistant.
No outcome was formally measured.
Read the case →I take a small number of advisory and build engagements each year.