Aditya Gaikwad
I build agentic AI systems and ship them as full-stack apps, from LLM tool-calling and rendering pipelines to the UI, APIs, and cloud deployment.
I'm a 2026 Computer Science graduate focused on agentic AI, building LLM-driven systems where natural language turns into schema-validated actions with function calling, self-correction, and real tool execution. My projects Atelier and Clavis are both agentic AI applications shipped end-to-end.
I pair that with strong full-stack engineering, turning AI ideas into apps with TypeScript, React, Next.js, Node.js, Python, and FastAPI. I'm comfortable across the whole lifecycle: LLM integration, API design, secure auth, testing, and cloud deployment.
Projects / Work
Clavis: AI Agent Command Center
An AI agent command center that acts across Gmail, GitHub, Slack & Discord, with a risk engine that gates high-risk actions behind step-up auth.
Atelier: Agentic Video Editor
A full-stack web video editor driven by an LLM agent. Natural-language commands become schema-validated editing operations with self-correction.
Image-Based Forest Analysis & Optimal Path Computation
An IEEE-accepted computer-vision system: an end-to-end ML pipeline with a benchmarked YOLOv8 model for forested-area analysis.
Experience
B.Tech, Computer Science & Engineering
Shri Ramdeobaba College of Engineering and Management, Nagpur
CGPA 8.11 / 10. Focus on application architecture, full-stack development, and machine learning.
Diploma in Computer Science
Government Polytechnic, Nagpur
Graduated with 93%, building the fundamentals in programming and systems.
IEEE Publication
Research paper accepted and published at an IEEE International Conference.
Swiggy Builders Club
Idea accepted into the Swiggy Builders Club.
Inventory App in Production
Built an inventory management app adopted and used by an IT company (during diploma).
HackerRank Security Finding
Identified a security vulnerability in the HackerRank desktop platform during testing.
Cluely Bug Bounty
Reported a security bug in Cluely through its bug bounty program.
Publication
Image-Based System for Forested Area Analysis and Optimal Path Computation
Environmental management depends a lot on monitoring the forested areas, but being able to assess them accurately is hard. Large-scale field experiments or special equipment is typically needed for such surveys. In this study, an image-based prototype is developed which is capable of passively detecting growth and development from a distance. Trees are being assessed to identify optimum paths and to assess the extent of the vegetation cover, making the technology for automatic forest analysis from satellite imagery a technology of great potential. The system estimates vegetated cover, extracts tree crowns using YOLOv8, and takes the least-cost route using a density-aware cost function and Dijkstra's algorithm. This prototype is optimized for being light, modular, and scalable for a rapid initial forest evaluation.
