B.Tech IT Undergrad · Manipal Institute of Technology, Bengaluru
Building in: LLM Integration · Geospatial AI · Full-Stack · ML Systems
About & Skills
B.Tech Information Technology undergraduate at Manipal Institute of Technology, Bengaluru (graduating 2027), with hands-on production experience in full-stack development, applied ML, and AI systems.
Currently on an AI internship at the National Informatics Centre, Govt of India — building geospatial AI with LLaMA 3.3-70B and ArcGIS for natural language district analysis, census anomaly detection, and automated PDF reporting.
Experience
Selected Work
Full-stack AI platform ingesting business SOPs → automated process intelligence → interactive flowcharts via ReactFlow. Single upload triggers: PDF parse → LLM analysis → graph extraction → 7 KPI benchmarking → consulting-grade PDF report.
Detects & classifies power disturbances — sag, swell, harmonics, transients — with 99.16% accuracy. Voice AI assistant, real-time FFT dashboard, IEEE-aligned simulation.
Fine-tuned Qwen2.5-Coder-7B via QLoRA on an 805-example, 15-category benchmark, reaching 4.58/5 accuracy on an LLM-judge harness — end to end from data collection to Hugging Face deployment.
Full-stack real-time traffic system streaming updates across 24 zones every 2 seconds, with Dijkstra routing over a 153K-node OSM graph and a 27x congestion-engine speedup.
Python execution recorder/replayer library published to PyPI, with a self-contained HTML report generator — timeline, call-tree, and memory visualizations.
Skills Map
AI-Powered
// Powered by Claude · Ask anything about my work, skills, or background
Interactive
// Move your cursor to connect the dots
Journey
Joined Manipal Institute of Technology's Information Technology program — began building foundations in data structures, algorithms, OOP, and web development. Found a passion for building things that actually work in the real world.
Built StateLens, a Python execution recorder/replayer published to PyPI, and began fine-tuning language models on early versions of what became Kosmo. First real taste of shipping data-driven tools from scratch.
Deployed the company's production website end-to-end. Integrated Google Sheets API for real-time lead capture, implemented client-side security, and resolved critical performance bottlenecks. First real production deployment experience.
Building a GIS-based AI chatbot for district geo-analysis at a national scale. LLaMA 3.3-70B + ArcGIS + NL-to-SQL + Z-score anomaly detection across 28 spatial layers. The biggest real-world AI system I have built so far.
Recognition
200+ Data Structures & Algorithms problems solved on LeetCode, with ongoing activity in competitive programming.
View Profile ↗Co-authored "Psychological and Demographic Factors Influencing Phishing Susceptibility are Underexplored in the Design of Intelligent Detection and Training Systems" — submitted to CISCOM 2026.
Received strong judge feedback for GridSense AI's real-world power-quality diagnostic approach.
Let's Connect
Seeking internship and research roles in software engineering, full-stack development, and applied machine learning. Open to collaborations and professional connections.