A record of the work, research, learning and achievements behind the machines — every drawer opens onto its own verified sheet.A record of the work, research, learning and achievements behind the machines — every drawer opens onto its own verified sheet.
I trained two models from scratch and fine-tuned three pretrained ones — all for NLP and generative-AI work. Most of the job was making experiments repeatable: data preprocessing, tokenization, hyperparameter sweeps.
The other half was evaluation — precision, recall, F1, baselines, and reading prediction errors closely enough to know what to fix next.
In between: transformer architecture experiments — attention mechanisms, inference workflows — and the evaluation pipelines that held all of it together.
I've shipped eight-plus client websites end-to-end — from brief to launch.
I cleared GATE 2026 — both papers: Computer Science and Data Science & Artificial Intelligence. All India Rank ≈6,000.
GATE 2027 prep is already on.
My English–Hindi translation Transformer, built from scratch, is published — ICTCS 2025, Springer Nature's LNNS series (vol. 1877), with an extended preprint on Research Square.