Fire-Archy Project

Awarded first place for the project Fire-Archy: a three-part solution that:

• Included a mobile application backed by a NASA-patented decision tree for wildfire evacuation protocols
• Designed universal car filter using NASA patents for emergency conditions
• Incorporated NASA FIRMS data for predictive wildfire monitoring

Recognized under the NASA T2U initiative for integrating NASA IP into real-world solutions.

• Optimized vision-language models (Florence, BioClip) for 1M+ biological data points analysis
• Accelerated embedding workflows by 30% through GPU optimization of cosine similarity algorithms
• Designed zero-shot evaluation achieving 92% genus-level accuracy in bird species classification

• Delivered 50+ lectures on React/JS to 350+ students, improving comprehension by 15%
• Designed modular full-stack project workflows adopted by 80% of students
• Reduced project lifecycle bottlenecks by 25% through curriculum alignment

• Engineered data pipeline for 26+ Indian dialects (10K+ hours audio)
• Improved ASR accuracy by 15% through feature extraction optimization
• Fine-tuned Whisper LLM (769M params) achieving 90% F1-score

• Curated 1,000+ medical dataset enhancing language interpretation by 25%
• Fine-tuned LLM on 50K+ patient notes reducing annotation time by 30%
• Improved clinical note accuracy by 20% through EHR integration

• Increased user engagement by 40% through responsive design improvements
• Boosted conversion rates by 25% via data-driven UI enhancements
• Implemented feedback system raising satisfaction scores by 15%