Data Science @ RIT (4.0 GPA) | AI/ML @ Petrabytes | Ex-L&T | BITS-Pilani | MIT-Manipal | Building AI Products on AWS
Transformer-based ReID across aerial-ground domains.
- ViT/VDT architecture on CARGO & Market1501
- Custom synthetic dataset with Blender-MATLAB pipeline (100K+ image pairs)
- Trained with SageMaker GPU; achieved 11% improvement over CNNs
📌 Poster:
Summarize web content, PDFs, and images using 10+ AWS Services like Bedrock and Kendra
- Claude + Kendra for text RAG, Rekognition for OCR/image AI
- One-click deploy using Terraform & GitHub Actions
- End-to-end AWS: Cognito, API Gateway, Lambda, S3
📌 Architecture:
PPO and DQN agents on OpenAI Gym CarRacing environment.
- Shaped reward design + decay schedule tuning
- 2M+ episode training led to 40% faster lap times
- Featured in Agentic AI Innovation Challenge
📌 Certificate:
- 🧩 Rubik’s Cube Speed Solver: Sub-30s 3x3 average; also mastered 4x4 and 5x5 cubes
- ⚽ MIT Football League Head: Organized trials and structured a 250+ player league system with live ops
- 🎤 Research Talks & Hackathons: Presented posters and gave a few research talks at various RIT bootcamps in Rochester, NY
- 🧠 Cognitive Flow Junkie: I like solving bugs like I solve cubes—algorithmically and under pressure


