I am a Computer Science Engineering student at Vishwakarma Institute of Technology (VIT), Pune, specializing in Artificial Intelligence & Machine Learning (Expected May 2027).
I build high-performance backend systems and scale AI pipelines for real-world applications. My work focuses on distributed architectures, fault-tolerant system design, and sub-second latency optimization.
- Performance Driven: Engineered production-grade systems serving 1,000+ users with <450ms latency.
- AI/ML Focus: Experienced in LLM orchestration, RAG pipelines, and model fine-tuning/evaluation.
- Academics & Research: Maintaining a 8.39 CGPA with 2 IEEE publications in AI-driven systems.
|
|
|
|
- CodeSpyder Technologies | GenAI Intern | Feb 2026 – Present
- Managing a team of 6 interns developing a Learning Management System (LMS) scaling to 1,000+ users and 100+ concurrent sessions.
- Designed a FastAPI & Next.js backend with PostgreSQL (SQLAlchemy) via Supabase; engineered API Gateways & Rate Limiting services for ~200ms CRUD latency.
- Implemented Llama 70B model for AI-enabled mock interviews, tuning error handling to reduce AI latency to 800–1000ms.
- CanSpirit Artificial Intelligence | Software Engineering Intern - AI | Aug 2025 – Feb 2026
- Developed an Emotion Analytics Detection system for disabled students, analyzing 1,000 videos at 5 FPS.
- Implemented an Extra Trees algorithm achieving 88% accuracy, outperforming Deep Learning baselines, KNN (78%), and Random Forest (83%).
- Ensured consistent 800–1000ms API latency through structured pipeline logging and monitoring.
|
Real-time conversational AI screening using WebSockets, Groq (Llama 70B), TTS/STT, Monaco Editor, and Redis context caching.
Impact: Scaled to handle 50 concurrent WebSocket connections; automated reporting for 100+ users. |
Multi-agent pipeline (planner–executor) using FastAPI, JWT, Google OAuth, and GPT-3.5 Turbo to generate 5–10 slide decks in <10 seconds with dynamic media.
Impact: Reduced manual presentation creation effort by 80% with real-time generations. |
|
RAG pipeline embedding IPC text using Hugging Face (text-embedding-3) into FAISS, with MS MARCO Cross-Encoder reranking for legal precedent search.
Impact: Verified and measured retrieval effectiveness using Precision@k, Recall@k, and MRR. |
Deep learning computer vision system for automated brain tumor detection from MRI scans using transfer learning with EfficientNet.
Impact: Automated screening process of complex medical image data with low classification latency. |
- Research Publications: Contributed to 2 IEEE publications in conferences focusing on applications of AI and data-driven systems.
- Problem Solving: Solved 250+ LeetCode problems with a 1500+ contest rating (strong Data Structures & Algorithms foundation).
- Leadership: Managed a team of 10+ members for Vishwaconclave (stage, light, audio setups) and led 25 coordinators achieving 3,000+ registrations during SWDC's Matadhikar campaign.
|
|
|
|
|
|