Projects
Things I've built
Products, research prototypes, and passion builds. From LLM simulation engines to neuromorphic hardware models.
Pulp AI
A simulation platform that lets product teams test copy, features, and pricing against AI-generated customer personas, shipping decisions in minutes instead of weeks. Built on FastAPI, Postgres (pgvector), Redis, and Docker.
Prediction
Dominant cohorts
Clustered from reasoning outputs across all simulated personas
Edvo
An adaptive assessment co-pilot that targets each student's mastery gaps in real time via an LLM judge loop over a skill graph. Questions are selected from a hand-curated bank with LLM fallback generation. Field-piloted across after-school learning centers.
Skill Graph
Adaptive Question
LLM JUDGECurrent target skill
Quadratic Equations
AI REASONING
Student shows mastery gaps in factoring. Curated bank match found. LLM fallback not triggered.
Primeplay Rentals
A peer-to-peer rental marketplace for high-value scuba gear across India's dive communities. Validated via a no-code MVP built on a landing page, WhatsApp coordination, and UPI payments before committing to a full build.
Research & Academic
Research projects
CPI Forecasting across Indian Markets
Scraped ~1.2M CPI price points from 300+ Indian markets and trained LSTM models that reduced 30-day-ahead MAE by 18% versus ARIMA, revealing inflation hot-spots across eight economic zones.
Optimizing Pricing Strategies for Online Marketplace Sellers
Trained a 2-layer GRU on 120K marketplace listings to estimate demand curves and recommend launch prices. Lifted expected seller profit by 19% and cut MAE by 8% versus a GBDT baseline. Advised by Prof. Aswini K Mishra, BITS Pilani.
Thalamo-Cortico-Thalamic Relay on SpiNNaker
Modelled the TCT relay system on the SpiNNaker neuromorphic platform for low-cost, low-power cortical signalling. Built under Prof. Basabdatta Sen Bhattacharya at BITS Pilani.
Multi-Agent Reinforcement Learning for Cooperative Hunting
Built a predator–prey game with Cooperate/Kill actions and trained agents with Nash-Q learning. Demonstrated that learned policies converge to the game's Nash equilibrium, confirming stable cooperative dynamics.
Neural Machine Translation: Verb Dependencies
Built a customised Seq2Seq variant to improve NMT translation accuracy by reducing the proximity in verb phrases between source and target languages. Research conducted at IIIT Hyderabad.