Choose your track
Track 1: Predictive Intelligence (Analytics)
Focus. Use machine learning and predictive modelling to anticipate consumer actions, protect revenue and sharpen commercial decisions. Move marketing from reactive reporting to forward- looking intelligence.
Scenario A
Dynamic margin and price optimization
Friction: Multi-brand retailers can’t adjust prices fast enough without losing margin or volume to rival marketplaces.
Example build: An ML pricing pipeline tracking stock velocity, price elasticity, competitor price indices and holding costs to set real-time price updates.
Scenario B
Predictive churn and CLTV at risk
Friction: Subscription and SaaS brands spot disengagement only after cancellation, making win-back costly.
Example build: A behavioral model using app telemetry, login frequency and support tickets to flag at-risk accounts 30 to 60 days early, compute CLTV at risk and trigger interventions.
Open innovation: Find any bottleneck where predicting a future user state, transaction or demand curve gives marketing a clear edge.
Track 2: Agentic Marketing (Autonomous Workflows)
Focus. Go beyond chatbots. Build autonomous agents that plan, call tools and APIs, and check their own output, with a human in the loop.
Scenario AAutonomous B2B prospecting and lead enrichment
Friction: Sales teams lose hundreds of hours monthly researching companies, verifying contacts and writing cold emails.
Example build: Agent 1 crawls public company signals, Agent 2 scores intent and ICP fit, Agent 3 drafts personalized multi-touch sequences for human approval.
Scenario BBrand reputation and social escalation agent
Friction: Mentions on X, Reddit and review portals spike during launches and outages and overwhelm community teams.
Example build: An agent pipeline that filters noise, scores urgency and brand risk, drafts on-brand replies and routes serious incidents to CRM ticketing.
Track 3: Applied MarTech Deployment (Vision and NLP)
Focus. Apply computer vision, multimodal processing and NLP to images, video, speech and regional languages, bridging offline consumer environments and digital intelligence.
Scenario A
Computer vision for retail shelf monitoring
Friction: CPG brands pay for premium shelf space but can’t automatically verify compliance or out-of-stock SKUs.
Example build: An edge-friendly detection pipeline reading smartphone shelf photos, separating own products from competitors, calculating share-of-shelf and raising compliance alerts.
Scenario B
Multilingual and code-mixed sentiment decoder
Friction: Standard sentiment models break on Hinglish, Telugish, Tanglish and other code-mixed content.
Example build: A domain-adapted NLP pipeline detecting intent, aspect-level satisfaction (price, quality, delivery) and regional dissatisfaction.
Bring your own problem.
The scenarios above are illustrative. Propose real bottlenecks from retail, e-commerce, SaaS, BFSI or consumer goods, as long as they fit a track and pass the P.A.V. test.
JUDGING
The P.A.V. Standard
Every entry and final prototype is
judged on three pillars.
Pragmatic
Is the architecture technically sound, scalable and viable with current ML, agentic or data stacks?
Actionable
Can an enterprise marketing team deploy and run it within reasonable workflows, time and budget?
Valuable
Does it remove a high-friction pain point or move churn, CAC or revenue in a measurable way?
Two mentors on every team
Institutional Mentor
Faculty nominated from the participant’s
own college
- Acts as primary academic advisor
and day-to-day guide - Tracks technical milestones on campus
- Co-guides engineering with VJIM faculty
- Receives formal recognition from VJIM
and an invitation to the Marketing Conclave
VJIM Faculty Mentor
Assigned by VJIM after shortlisting
- Sharpens commercial viability and
enterprise marketing context - Validates against the P.A.V. standard
- Prepares the team for the finale pitch
and jury defense
How the challenge unfolds
Abstract submission
150-word problem definition, 200-word architecture breakdown and Institutional Mentor details.
Shortlisting and pairing
Desk review by the VJIM evaluation panel; shortlisted teams are paired with a VJIM Faculty Mentor.
Development sprint
Code build, pipeline testing, architecture coaching and presentation coaching with both mentors.
Grand finale pitch
10 Dec 2026. Live stage pitch and jury defense before the VJIM faculty council, VC and investor panel.
Winners announced
11 Dec 2026. Announced along with the Marketing Conclave; prizes presented by industry leaders.
Who can take part
Teams of 2 to 4 undergraduate
or postgraduate students from Engineering,
Technology, Data Science or Management streams.
One Institutional Mentor
from their college.
Original work only.
Open-source libraries, foundation models
and public APIs are allowed, but business
logic, workflow design and pipeline code
must be written for this challenge.
What to submit
Stage 1: Registration and abstract
- Team profile with institution and mentor details
- Track selection and problem statement
- 150-word solution abstract
- 200-word technical architecture with diagram
Stage 2: Finalist sprint
- Documented GitHub repository with
dependency files and setup instructions - Working prototype: deployed app, API endpoint or live terminal demo. Static mockups and slide-only entries are disqualified.
- Executive pitch deck, max 10 slides