01 / About
I am a second-year B.Tech IT student at the Indian Institute of Information Technology, Allahabad.
My core technical interests are in Machine Learning, Computer Vision, and NLP. I am pursuing a career in ML/Data Science, focused on building efficient infrastructure and agentic systems.
Open to research roles at labs and applied ML positions at companies.
02 / Experience
Research Intern
Centre for Intelligent Robotics (CIR), IIIT Allahabad
Leading research on query-conditioned Spatio-Temporal Transformers and Active Inference for world-model consistency.
Member
Geekhaven AI/ML Wing, IIIT Allahabad
Collaborating on department-wide AI/ML workshops and open-source mentoring.
03 / Papers
QueST: Persistent Queries as Semantic Monitors for Drift Suppression in Long-Horizon Tracking
Accepted at ICLR Workshop
Architected a novel transformer model to maintain tracking consistency in high-occlusion articulated motion scenarios.
04 / Projects
FinGraph: Multi-Agent Financial Intelligence System (March 2026) →
Developed a multi-agent orchestration pipeline for investment analysis and risk assessment (Top 10 at III 5.0 Hackathon). Engineered a 4-node acyclic graph (Market Data Fetcher → Technical Analyst → Sentiment Miner → Risk Auditor) powered by Tavily API and yfinance. Built a Stock Network Analyzer with NetworkX and Streamlit to visualize Pearson and lagged correlations.
Matsya AI: Intelligent Catch Analysis →
Engineered a Perception-to-Action pipeline using YOLO (detection) and ResNet (classification) to extract biological insights from fish imagery. Implemented a three-path weight estimation pipeline combining regression models, biological formulas, and contextual reasoning. Built an Agentic AI 'Brain' that analyzes catch history, weather, and market prices to recommend optimal selling strategies.
Narrative Consistency Engine →
Engineered a sophisticated local inference pipeline leveraging Qwen 2.5 7B with custom-trained Baby Dragon Hatchling (BDH) embeddings for semantic coherence analysis. Integrated deep learning architectures with XGBoost ensemble classifiers to detect cross-temporal narrative inconsistencies across parallel story arcs with high precision.
Eudexa: All-in-One Law Firm Companion →
Architected a comprehensive legal AI platform featuring Retrieval-Augmented Generation across 1,000+ unstructured documents with sub-500ms latency. Integrated intelligent meeting scheduling, automated risk-assessment calls to clients, and proactive case monitoring—transforming traditional law firm operations into a unified AI-driven workflow.
SmartKisaan: AI-Powered Agricultural Intelligence System →
Developed a full-stack mobile platform combining a custom-trained deep learning model for real-time crop disease detection with Google Gemini-enhanced remediation insights. Features include JWT-secured authentication, real-time weather integration via OpenWeatherMap, an agricultural marketplace, community forum, and an AI-powered farming assistant chatbot—all orchestrated through a scalable FastAPI backend with PostgreSQL.
More projects will be added soon
05 / Blogs
BIOMASS prediction from Image
Coming soon
How to start with ML as a fresher
Coming soon