✶ Product Management · AI Systems · Campus-scale Operations

Product systems.

I’m Aditya - a BITS Pilani student who turns messy operational problems into roadmaps, MVPs, dashboards and growth loops with visible business outcomes. I built and shipped Interviehire, an autonomous hiring platform, ran it for four months, generated revenue, then closed it. Now I’m interning as a Forward Deployed Engineer at Urumi.ai.

See Interviehire ↗ Resume ↓
Portrait of Aditya Pratap Rana
Aditya Pratap Rana · BITS Pilani ’27
5,000+users served across product & ops systems
91Cr+contract secured through demand validation
29Crbudget managed as mess society president
98%accuracy in a stellar-classification ML pipeline
001 — Flagship Venture

✶ Autonomous Hiring · Zero Human Intervention

Interviehire

Companies shared a requirement. Interviehire delivered verified hires in days - not months.

  • Built & shipped
  • Ran for 4 months
  • Generated revenue
  • Closed

Interviehire ran the entire hiring funnel end to end. We sourced candidates, shortlisted resumes through our in-house screening engine, and ran automated recruiter screening. Everyone who passed met LINA, our AI interviewer avatar, for a dynamic functional interview - with live eye-tracking that flagged anyone reading their answers, plagiarism checks on every response, and full proctoring. Each candidate left with a detailed report spanning everything from resume to interview. No humans in the loop - companies shared the role, we delivered the hires.

0humans in the hiring loop
100%of answers plagiarism-checked
6automated stages, sourcing to report
Daysfrom requirement to verified hire
  1. 01

    Sourcing

    Reached the right candidates at scale for the exact role a company needed filled.

  2. 02

    Screening Engine

    Our in-house engine shortlisted resumes against role-specific signals - instantly.

  3. 03

    Recruiter Screen

    An automated recruiter screening round qualified candidates before they interviewed.

  4. 04

    LINA Interview

    Our AI avatar LINA ran a dynamic functional interview with live eye-tracking.

  5. 05

    Integrity Checks

    Every answer was plagiarism-checked, backed by full standard proctoring protocols.

  6. 06

    Candidate Report → Hire

    A detailed report per candidate - resume to interview - and the hire, in days.

Drag or swipe →
002 — Independent Builds

Shipped outside the classroom.

Three solo, end-to-end builds - real code, real data, real infrastructure.

Platform Engineering · Kubernetes

Store Orchestrator

A multi-tenant platform that provisions a fully working online store on Kubernetes from one click or one API call, in about two minutes. Every store gets its own namespace, database, quota, limits and network policy, and arrives with a seeded catalog and Cash on Delivery, so it can take an order straight away. The same Helm charts run on a laptop (Kind) and on a production VPS (k3s with Let’s Encrypt TLS) - only the values file changes.

  • Crash-safe by design. Kubernetes is the source of truth; a reconciler resumes any half-finished install or teardown after a restart.
  • Idempotent creates. Store ids derive from the request key, so retries and double clicks never create a second store - even across replicas.
  • Horizontally scaled. Leader election on a Lease, Postgres advisory locks for quotas, 2–5 API replicas behind an autoscaler.
  • Least privilege. Scoped RBAC plus a ValidatingAdmissionPolicy, so the API can only ever touch store namespaces.
  • Safe upgrades. One-store-at-a-time canary rollout with backup, smoke test and automatic rollback.
~2 minfrom click to a store that takes orders
19store types auto-detected for the catalog
1 callfull teardown, no leaked volumes
KubernetesHelmTypeScriptFastifyReactPostgrescert-managerPrometheusGitHub Actions
Full-Stack Build · Live Data

HiringMap

Map-first job discovery for India: open a map, zoom into a city, and see exactly who’s hiring there - click through and apply straight at the source. Built the ingestion pipeline that crawls ATS boards directly (Greenhouse, Lever, Ashby, Workday, Workable, Recruitee) plus a licensed aggregator, a PostGIS-backed map that rolls pins into city/state clusters and scatters them apart so no name gets hidden, and a FastAPI + Next.js stack backed by 460+ tests.

7,532+companies indexed
13,177+active job postings
202ATS boards crawled directly
FastAPIPostGISNext.jsIngestion pipelinemypy strict
View the code ↗
Mobile Prototype · Solo Build

Referly

A Hinge-style swipe, match and chat experience - for hiring instead of dating. Job seekers and recruiters/referrers swipe on each other; a mutual match triggers a real in-session chat thread. One role toggle reshapes Discover, Likes, Matches and Activity depending on which side you’re on. Built end to end on Expo/React Native, with a full onboarding flow, a gesture-driven swipe deck, and a Referly+ paywall screen.

7core flows, onboarding to profile
2roles, one toggle - seeker & referrer
0backend - fully client-side prototype
ExpoReact NativeReanimatedGesture HandlerTypeScript
View the code ↗
003 — Product Case Files

Framing, prioritization, measurable outcomes.

A portfolio built around real problems, sharp trade-offs, and numbers that moved.

0–1 Platform

SSMS Smart Mess Platform

Defined the roadmap and an A/B pilot for a mess app, using rating feedback to optimize menus and drive 15% higher attendance.

QR paymentsSQL validationFraud reduction
Retention Teardown

Flipkart Delivery Reliability

Diagnosed a 38% repeat-order decline, prioritized a reliability MVP with RICE 96, and linked fulfilment accuracy to a projected 18% D30 retention lift.

North StarRICED30 retention
MVP + Monetization

Stride Accountability App

Researched 80–90% early churn, scoped MVP activation loops, and projected INR 45L Year-1 revenue through subscription monetization.

User researchAWS architecturePricing
004 — Machine Learning Lab

Technical work that sharpens product intuition.

Hands-on ML projects behind the way I reason about AI-powered workflows.

Research · Astrophysics × ML

Predicting Star Formation and Stellar Type

Research on predicting star formation and the type of a star with machine learning. The model is trained on synthetic stellar data and learns how a star’s physical properties relate to one another: give it any three known properties of a star and it predicts the other two. Automated preprocessing, feature engineering and recursive modeling in Pandas and NumPy fill in missing stellar attributes, and the classifier identifies star types at 98% accuracy.

3 known properties Model trained on synthetic data 2 predicted properties + star type
PythonPandasNumPySynthetic dataRegressionClassification98% accuracy
Research · Biomedical Sensing × ML

Automated Regression Pipeline for S-Parameter-Based Prediction

Estimates blood glucose non-invasively from microwave S-parameter readings (S11, or two channels s1/s2), following the method of recent antenna-sensor research (IEEE TIM, 2025). The pipeline standardises the inputs, selects features with recursive feature elimination, then trains and compares linear regression, random forest and gradient boosting on an 80/20 holdout, with hyperparameters tuned by 5-fold cross-validation on MARD. A Streamlit app turns a new reading into a glucose estimate in real time.

S-parameter reading Scale, RFE, LR / RF / GBDT Glucose estimate
Pythonscikit-learnPandasStreamlitRFEGradient boostingR² · RMSE · MARD
005 — Operating Roles

Budgets, stakeholders, revenue, thousands of users.

Leadership where the outcomes were owned end to end.

President · Society for Student Mess Services

Managed a 29Cr budget, improved adoption across a 5,000+ user base, drove 60% growth, and built a proposal securing a 3-year 91Cr+ contract with 70L+ investments.

Coordinator · Hindi Activities Society

Reduced costs by 23%, secured 30L in partnerships, owned P&L for the technical festival, and led an 18+ member team across 10,000+ attendees.

Secretary · Radioaktiv, BITS Pilani

Generated 10L+ revenue at 67% ROI, launched four programs, and directed a two-tier team of 50+ members across 3,000+ participants.

Campus Director · HULT Prize Foundation

Mentored startups to the regional level, helped one secure 1Cr funding at a $1M valuation, and built a repeatable GTM playbook for a 100+ cohort.

006 — Experience
Current Forward Deployed Engineer Intern · Urumi.ai

Deploying an AI-run commerce platform for D2C brands

Urumi.ai hosts and runs WooCommerce stores for direct-to-consumer brands, with AI agents that build features, guard revenue at checkout and answer analytics questions - overseen by engineers who built WooCommerce core, with $60M+ in GMV on the platform. I work directly with brand teams on their live stores:

  • Migrations. Move brands’ stores onto the platform and get them live.
  • Integrations. Connect each store to the third-party tools and services the brand runs on.
  • Features and automations. Build custom store features and automate repetitive operations work.
  • AI agents and analytics. Set up the platform’s AI agents and MCP connections, and turn store data into answers the brand can act on.
WooCommercePHPMCPAI agentsAnalyticsIntegrations
Summer Intern · BARC · May–Jul 2024

Transmission efficiency study

Identified system inefficiencies and recommended optimizations that improved power transmission by 15% and reduced costs by 12% - earning a Letter of Recommendation from the BARC(V) Project Director.

007 — Now Routing

Let’s build something worth hiring for.

Open to product internships, AI product builds, and operations-heavy PM roles.

Email me LinkedIn