Vikranth Babu T

London · Open to Applied AI leadership & senior technical ownership

Applied AI that earns its place in production

I’m applied AI and Data Science practitioner. For over a decade, my work has lived where machine learning drives high-stakes decisions—optimizing hotel bids, personalizing shopping feeds, and vetting production LLMs. Whether stepping in as an applied AI leader, tech lead, or architect, I stay close enough to the models, evaluations, and code to ensure every system delivers measurable value with speed and efficiency.

Get in touch Download CV See my work

  • LLMs & AI Agents
  • LLM Evaluation
  • Bidding & Optimisation
  • Ranking, Recommendations & Personalisation
  • Experimentation
  • Marketing Science

How I like to work

  1. FrameThe decision first, then the model
  2. MeasureNorth star, guardrail, launch bar
  3. EvaluateOffline first, calibrated to people
  4. TestLive A/B, decision rules agreed up front
  5. DecideScale it, change it, or stop it
  6. Own the outcomeLive, measured, improving
11+ yearsin applied ML
Travel & e‑CommerceExpedia, Skyscanner, eBay
Tech LeaderTeams in the US, UK and China
BuilderHands-on, from PRD to code
Digital AdvertisingPaid search, shopping, affiliate, social

What I do

Where I add value

Applied science that changes a business decision, backed by the evaluation to prove it worked.

Decisioning & ML systems

Models that spend real budgets and rank real inventory.

  • Bidding & budget allocation: Optimizing high-stakes financial levers.
  • Ranking, recommendations & personalization: Shaping what users see and buy at scale.
  • Experimentation: Rigorous A/B and Bayesian testing frameworks.

LLM systems & evaluation

Generative AI that a business can actually stand behind.

  • Production LLMs: Fine-tuned products owned live in production.
  • Rigorous evaluation: Rubrics, automated LLM judges, and human review loops.
  • Architecture: Responsible AI, RAG, and multi-agent design.

Leading applied science

Driving initiatives from the initial problem statement to a live result.

  • Technical product ownership: Problem framing, PRDs, and strict launch criteria.
  • Hands-on leadership: Directing lean, distributed teams while staying deep in the details.
  • Cross-functional alignment: Bridging product, engineering, and analytics.

How I work

From the decision to a measured result

01

Frame

What decision does this serve?

  • Determine if model is the right lever
  • Name the target outcome before picking the method
02

Measure

Agree how we measure it

North star, guardrail and launch bar, agreed before anything is built.

03

Build and Evaluate

Test rigorously

  • Offline evaluation first, calibrated to human judgment
  • Live test with decision rules set up front
04

Decide

Own the outcome

Own the readout and the call that follows: scale it, change it, or stop it.

Selected Work

All projects and case studies

AI Listing Descriptions

2025

Customer-facing LLM that drafts listing descriptions for sellers in the US, UK and Germany, served on vLLM.

  • Owned the product in production with MLOps
  • Three regional models consolidated into one multilingual model
LLM EvalsMLOpsProduct management
eBay

AI-powered Personalized Banner Copy

2025

LLM system writing personalised “Sell one like this” messages for customer groups across three markets.

  • LLM judge matched human reviewers at ~90% precision and recall
  • PRD to live A/B test in three months
LLM EvalsPrompt EnggHands-on builderSolution Architect
eBay

Promoted Offsite Ads

2022–23

Bidding science and architecture letting sellers promote their listings on Google, with one campaign per seller.

  • $53M incremental GMV, ~20k seller campaigns
BiddingSolution Architect
eBay

Hotel Ads Bidding System

2026

Designed and built a new Google Hotel Ads bidding system, adapting the architecture to platform constraints.

  • Concept to 3 live A/B tests as sole builder
BiddingSole builder
Skyscanner

Listing Title Rewrite

2022

Model rewriting listing titles across the whole Google Shopping feed, built on existing entity recognition.

  • +500K daily impressions
NLPTeam lead
eBay

Partner Network Recommendations

2024

Recommendation model for a new eBay Partner Network programme, taken through four iterations from pilot to MVP.

  • Zero production incidents; patent application accepted for filing
RecsysTeam lead
eBay

Experience

My Background

  1. 2026 – nowLondon (UK)
    Vihive Labs Ltd
    • Independent AI consultancy
    • Clients: Skyscanner UK Ltd
    • Redesigned in-house bidding system
    • Brief to 3 live A/B tests
    Company directorSolution designerSole builder
  2. 2021 – 2026London (UK)
    eBay
    • Performance Marketing & Selling Science (Paid Search, Shopping, Affiliates, Social)
    • AI Applications Science
    • 20+ ML initiatives from roadmap to production
    • GenAI applications (Listing descriptions, Personalised messaging, Copilot assistant design)
    Team leadAgile tech leadProduct managementHands-on builder
  3. 2015 – 2021Seattle (US), London (UK)
    Expedia Group
    • Bidding science and operations
    • Marketing ML platform: Operations and management
    • Cross-channel, Cross-partner science initiatives
    • Meta-search, SEM, Search Compare channels
    • Experiment Design and setup
    • (COVID era) cancellation prediction model
    Hands-on builderTech leadProduct ownerLine manager
  4. 2010 – 2014
    • T-Mobile · Statistician Intern (2014)
    • IIM Ahmedabad · Research Associate (2012–13)
    • Juniper Networks · Software Engineer II (2011–12)
    • Avaya Labs · Research Intern (2010)
  5. 2007 – 2015
    Education
    • University of Washington, Seattle · M.S., Industrial & Systems Engineering (2015)
    • IIT Madras · B.Tech., Electrical Engineering, minor in Operations Research (2011)

Who I am

About me

Vikranth Babu T
Vikranth Babu T

Applied AI & ML leader · London

I’ve spent more than eleven years working where machine learning directly drives business decisions—determining what to bid for a hotel click, which listing a shopper sees, or whether an LLM’s output is safe to show a customer.

I started in software engineering and operations research, which is why I care just as much about the commercial decision a system serves as the model itself. I do my best work leading lean teams on deep problems that matter to the business, and I still review the architecture, the evaluations and the code myself.

Alongside that, I consult independently through Vihive Labs and build with multi-agent systems and LLM evaluation. Outside work: cricket with Croydon Cricket Club, travel and hiking.

PythonSQLPySparkPyTorchXGBoostLLMs & evaluationvLLMRAGMulti-agent designA/B & Bayesian testingDatabricksAWS
Publications & patents

Building an applied AI team?

If you’re hiring for applied science or ML leadership, or want a second opinion on an AI problem, I’d be glad to talk.

Email me LinkedIn Download CV