SWARNIM
TRIPATHI
Code is easy. Reliability is hard. I build safety-first agents, adaptive learning engines, and backends that handle chaos with mathematical precision.
Core Intelligence
Infrastructure
Languages
Contributed PR #32046 aligning CUDA 13 installation instructions for improved developer experience. Updated documentation to reflect the latest NVIDIA ecosystem changes.
Machine Learning Engineer Intern
- Hybrid Ranking Engine: Enabled real-time matching of 16,000+ garments by architecting a hybrid ranking engine using CLIP embeddings and FAISS, optimizing for high-throughput vector search.
- Cold-Start Solution: Solved the "cold-start" problem for new users by formulating context-aware User Vectors, combining facial/physical attributes with rule-based filtering logic.
- Latency Optimization: Achieved sub-100ms inference latency by partnering with backend teams to integrate optimized decision logic into FastAPI endpoints.
Aura
Hybrid glucose forecasting system combining personalized LSTM models with event-based adjustments. Features a Deep Q-Network (DQN) agent that simulates optimal insulin dosing strategies.
:: View Source Live DemoPromptimus
AI prompt optimizer fine-tuning a 220M parameter T5 model with LoRA & RAG. Solves intent preservation using a custom 200k dataset and constrained beam search.
:: View SourceLearnBuddy
Adaptive learning engine merging Multi-Armed Bandits (MAB) with heuristic rules. Dynamically adjusts difficulty based on real-time confidence scores and success rates.
:: View SourceAnxiety Predictor
Anxiety level predictor using SGDRegressor on lifestyle datasets. Features robust preprocessing with missing value imputation and advanced encoding techniques.
:: View SourceStudyBoard
AI-powered command center for academic excellence featuring daily strategic briefings, adaptive scheduling, deep work timer with focus analytics, and gamified achievements. Built with Next.js 15, Shadcn UI, and Supabase.
:: View Source Live DemoGhost
Autonomous AI Test Agent for Python that generates context-aware pytest suites via AST analysis. Features a self-healing engine that auto-fixes failing tests and a "Judge" protocol to distinguish test bugs from code bugs.
:: View Source PyPIFueled by curiosity and caffeine.
CAF
FEINE
SYNTH
WAVE
SCI
FI
Why deploy me?
I don't just write code; I ship products. From setup to deployment, I handle the full lifecycle so you don't have to handhold.
I understand ROI. Every line of code is written with the business goal in mind—scalability, maintainability, and user value.
Systems break. I build them to recover. My experience with distributed systems means I design for failure from the start.