Alpit Kale

Procurement · Supply Chain · Applied AI

I design procurement and supply chain change, then build the tools that run it.

I’m Alpit, a Manager in a Big 4 supply chain and procurement practice in London, with six years across Big 4 and specialist consultancies. Alongside client work I build and run production software: LLM pipelines, data migrations, and a manufacturer’s nightly margin dashboard.

Selected work Download CV
alpitkale@gmail.com linkedin.com/in/alpitkale github.com/Alpit-K London, UK

About

These days my work runs on two tracks that feed each other. In London I’m a consultant in supply chain and procurement, where most of my time goes on turning tangled operational problems into software people actually use day to day.

Alongside the advisory work I build and run production software hands-on for a manufacturer in India — the systems a factory business relies on to see its own numbers and make decisions. The two feed each other. Real factory data shows me which ideas hold up in practice, and client work sets the standard I hold the software to.

Off the clock I DJ, dive, and photograph wildlife. I also built a pocket-sized drum machine to layer live drums over my own sets (how it came together →).

Journey

Four cities, one ship, fourteen countries at sea. Scroll — the route draws itself.

    Work

    In every role I’ve tried to own a problem from definition through build to adoption.

    2022 — Present London, UK

    Manager — Deloitte

    Supply chain & procurement transformation, AI-enabled delivery

    • Manager in the supply chain and procurement practice, leading AI-enabled transformation delivery for global pharma, FTSE 100 TMT, and UK telecoms clients.
    • Nominated AI Change Champion by senior leadership, with a firm-wide remit for AI adoption strategy and practitioner capability building; firm-wide Innovation Award nomination (2026).
    • Technical lead on Cascade, a Gen AI supply chain resilience prototype built for the firm’s internal innovation competition.
    • Delivered demand planning analytics (SQL, PowerBI) for a global technology client’s devices supply chain; mapped multi-tier supplier networks for an Australian enterprise; formulated a post-merger supplier strategy for a UK telecoms client.
    Procurement transformationCLM AI product developmentPython Stakeholder management
    Case study: AI Contract Inquiry → Case study: AI Intelligence Briefing → Case study: Cascade →
    2020 — 2022 London, UK

    Senior Associate Consultant — GEP Consulting

    Strategic sourcing, category management, process automation

    • Optimised the supplier base for a German pharmaceutical multinational to streamline buying and category management.
    • Supported PMO and strategic sourcing for a FTSE 100 firm delivering a national-scale public health programme.
    • Implemented robotic process automation to amend contracts at scale and maintained PowerBI reporting dashboards.
    Strategic sourcingSpend analytics Category strategyProcess automation
    2017 — 2019 London · Copenhagen · Hong Kong

    Earlier roles

    Investment research, communications, startup research

    • Investment Research Summer Analyst at a London wealth manager — emerging-market macro datasets and bottom-up equity valuations.
    • Digital communications at a global energy engineering firm in Copenhagen — 25+ published articles, 35% LinkedIn traffic growth.
    • Startup ecosystem research across Hong Kong and India.
    Investment researchCommunicationsMarket research

    Speaking & teaching

    1. Sep 2026 Demoed Cascade, with a colleague, to around 200 practitioners at a practice-wide AI session. It ran long on questions.
    2. Sep 2026 Helped run a practice hackathon, at a couple of hours’ notice.
    3. Jul 2026 A 45-minute internal talk on AI agents for consultants who advise clients on them, using my own builds as the worked examples.
    4. Jul 2026 First interviewee in an internal series on how practitioners build with AI, on the briefing agent.
    5. Autumn 2026 Co-presenting a client webinar on AI in supply chain planning: data foundations, tech stacks and a maturity assessment.

    Skills

    Functional

    • Procurement transformation
    • Supply chain strategy
    • S&OP & demand planning
    • Margin & cost analytics
    • Commodity & supplier intelligence
    • Contract lifecycle management

    AI & LLM

    • AI product development
    • RAG systems
    • Agentic workflows
    • Prompt engineering
    • Claude Code

    Technical & tooling

    • Python
    • SQL
    • N8N
    • RPA

    Education

    King’s College London
    BSc (Hons) Business Management

    Copenhagen Business School
    Finance & Economics

    Semester at Sea
    Global studies across 14 countries

    Projects

    Things I designed and built end to end, for clients, colleagues and a manufacturer. Most are in daily use.

    AI Intelligence Briefing
    Internal product 200+ practitioners

    AI Intelligence Briefing

    A two-model LLM pipeline that reads sixteen sources and ships a weekly strategic brief as branded HTML and PDF. Nominated for a firm-wide Innovation Award (2026).

    View case study →
    Client delivery

    AI Contract Inquiry Solution

    AI triage and answers for a global pharma client’s contract queries, inside its existing CLM workflow.

    ~40Kcases a year

    The problem

    Contract queries arrived in volume, every day, and each one pulled a procurement specialist away from real work. The answers lived in the contracts — but finding them was manual, repetitive, and slow.

    What I built

    An AI-enabled inquiry solution sitting inside the client’s existing CLM workflow: routing logic to triage queries, LLM pipelines to resolve them against contract data, and escalation paths for the cases that genuinely needed a human.

    The hard part

    Making it reliable. Handling ambiguous queries, getting the routing edge cases right, and earning enough trust from procurement colleagues that they actually stopped answering manually. The design constraint throughout: fit the existing workflow — no process change on the client side.

    Outcome

    Roughly 40,000 manual case responses eliminated per year, with the client team redirected to higher-value procurement work.

    Client delivery

    Contract Migration Programme

    Technical architecture for a FTSE 100 TMT client’s contract data migration, in UAT ahead of a December 2026 production load.

    ~5Kcontracts

    The problem

    Around 5,000 contracts needed to move into a new structure — with zero tolerance for data loss and a client team that had to approve every structural decision along the way.

    What I built

    Client-approved mapping workbooks are the single source of truth — a code generator turns them into the full SQL pipeline, so every transformation is signed off before it exists, and validation gates run before any batch is exported. No hand-edited SQL, paired with a daily governance cadence that kept the client team confident and unblocked.

    The hard part

    Data quality and stakeholder confidence. Migration programmes fail quietly — a field mapped wrong, a clause dropped — so the discipline was in validation at every stage and in keeping decision-making visible to the client rather than buried in the pipeline.

    Outcome

    The workbook design pays off in iteration: a change to a mapping produces a new load file in about a minute, with no SQL edited. The first load files cleared client review, with the remaining gaps traced to business inputs rather than the transformation. User acceptance testing is under way, and the production load is scheduled for December 2026. Presented at a partner-level showcase.

    Internal product

    Cascade — Supply Chain Resilience

    A Gen AI supply chain resilience tool for the firm’s innovation competition. Technical lead, team of three; before the panel in November 2026.

    ~200at the first demo

    The problem

    How exposed is a supply chain to disruption, and what should the organisation do about it? Cascade answers from a model of the network itself, using the firm’s resilience methodology rather than generic advice.

    What I built

    A Python pipeline end to end, with a React front end: a graph model of the supply network, scenario and scorecard generation driven by the firm’s methodology, and a maturity assessment on top. It runs on local models by default, with an optional path to a frontier model where a client’s data policy allows it.

    The hard part

    Getting a three-person team building, two of them senior to me. Six weeks of design documents produced agreement but little hands-on work; a rendered mock-up on synthetic data changed that in an afternoon, and an hour walking a colleague through the codebase did more than any task-allocation process.

    Where it stands

    The build is complete and demoed; the next constraint is how much of the firm’s material leadership releases into it. The competition panel is in November 2026.

    Built for myself

    The finished Pocket DJ Drum Machine

    Pocket DJ Drum Machine

    Built, boxed and playing
    Open sourceTeensy 4.1C++17
    View case study →
    DJ VJ — a still of the shards scene DJ VJ — a still of the rings scene DJ VJ — a still of the cells scene

    DJ VJ — Live Visuals

    V1 done, played at a party
    Open sourceWebGL2Web Audio
    View case study →

    How I work

    i

    Judge it by use

    Work is finished when the people it was built for use it every day. That’s why most of my projects end in a tool rather than a slide pack.

    ii

    One source of truth

    Most operational pain is the same data living in five places. Whether it’s 5,000 contracts or a factory’s ERP, the fix starts with agreeing which version is correct and generating everything else from it.

    iii

    Prototype in days, harden what survives

    A first version goes in front of real users within days. Whatever they keep using then gets tests, validation and a proper deployment.

    iv

    Never invent a number

    When the data can’t support a figure, the tool shows a blank and says why instead of filling in an estimate. It’s what let the margin dashboard pass a line-by-line check against audited accounts.

    Behind all of it is a personal knowledge vault: a daily note every day since April 2026 and a weekly review that hasn’t missed a week. Projects get scoped there before any code is written, and it’s where the facts on this site are checked before they go up.

    Play

    A 60-second pocket edition of PlannerOS, the supply chain planning simulation I built for consulting workshops. You’re Head of Supply Chain Planning. Four quarters, four calls to make. Your decisions reveal your planning archetype. Play the full simulation ↗

    Contact

    Always happy to talk procurement, supply chains, applied AI, music, or where to dive next. If you’re building something where domain depth meets hands-on AI delivery, even better.

    alpitkale@gmail.com
    linkedin.com/in/alpitkale github.com/Alpit-K Download CV (PDF) London, UK