📍 Bangalore, India | 🧠 ML Engineer | 🔬 Building AI from scratch — not just using it
Deep in the weeds of transformers, tokenizers, and training loops. I build models from the ground up to actually understand how they work. Currently grinding on NLP and computer vision while looking for my first ML/AI role.
🚗 ENFORE v2 — Safe Multi-Modal Autonomous Driving (MTP Thesis) Multi-modal perception pipeline for safe autonomous driving — fuses camera and sensor data using deep neural networks to enable real-time scene understanding, obstacle detection, and safety-aware decision making in complex environments
🤖 Small Language Model — Built From Scratch Decoder-only transformer (GPT-style): custom BPE tokenizer, multi-head self-attention, positional embeddings, nucleus sampling
🏥 Multimodal OCR Pipeline — Insurance Document Digitization CNN + metadata fusion model to classify scanned insurance IDs — PyTorch, 10-epoch training with LR scheduling
📈 Market Intelligence & Trend Analysis Engine Google Trends time-series engine across 20+ categories — seasonal decomposition and COVID-era demand analysis to guide fitness product-growth strategy
Transformers CNNs NLP BPE Tokenizer Multi-Head Attention BeautifulSoup
"I don't just call
.fit()— I build the pieces that make it work."