const lalitaditya = {
pronouns: "he/him",
location: "India 🇮🇳",
role: "AI/ML Researcher + Full Stack Dev",
university: "SRM University",
year: "B.Tech CS — 4th Semester",
currentlyBuilding: [
"🔬 Underwater Marine Litter Detection (Research Paper)",
"🏥 CXR-ReportGen — Auto Radiology from X-Rays",
"🧠 ScholarMind — RAG-powered Research Assistant"
],
stack: {
weapons: ["Python", "PyTorch", "React", "Next.js"],
ai: ["Computer Vision", "Transformers", "LoRA"],
backend: ["FastAPI", "Node.js", "MongoDB"],
learning: "MLOps + distributed training"
},
vibe: "shipping research > shipping hype",
mantra: "papers don't write themselves 💀",
askMe: "anything CV / deep learning / papers"
}; |
|
Comparison of Mask R-CNN vs Mask2Former on 8,000+ annotated underwater images across 10 categories. Features underwater-specific augmentation pipeline, differential learning rates, mixed precision training.
|
EfficientNet-B3 (visual encoder) + BioGPT (medical text decoder) with LoRA fine-tuning, trained on the IU X-Ray dataset. Generates clinical-grade findings and impressions.
|
Built with FastAPI + ChromaDB + sentence-transformers + Claude API + React. Semantic search over papers with source-grounded responses.
|
studying: Advanced Computer Vision
researching: Transformer-based segmentation
reading: Papers (arxiv.org my beloved)
building: Research portfolio for conferences
target: CIS 2026 → ADCIS 2026 → ICCAIML'26
mood: "the paper won't write itself" 😤 |
caffeine_intake: "dangerously unmonitored ☕"
debug_hours: "3 AM is my peak performance"
fave_language: "Python (don't tell JS)"
current_anime: "Naruto Shippuden 🌑"
superpower: "turning research papers into code"
weakness: "CSS centering div 💀" |
