I'm an AI/ML engineer and data science student focused on reliable, human-centered AI systems for healthcare, learning, and social good.
My work sits at the intersection of clinical data pipelines, privacy-preserving ML, audio deep learning, and human-in-the-loop AI. I care about systems that are useful in the real world: auditable, reproducible, local-first when privacy matters, and designed to augment human judgment instead of replacing it.
My path has been non-linear: startups, business, service, contemplative discipline, and now rigorous AI/data engineering. That path is the reason I care so much about high-agency execution, deep learning, and building technology that makes people more capable.
- Clinical and behavioral ML for neuro-diverse populations
- Human-in-the-loop workflows for research-grade AI systems
- Robust audio and NLP modeling under messy real-world conditions
- Local-first/private inference for sensitive data
- Long-term: AI systems that help humanity become healthier, wiser, and more capable
| Project | What it demonstrates |
|---|---|
| Clinical ASD Screening Ensemble | Leakage-safe clinical ML, calibrated ensembles, reproducible inference |
| Touch Data Quality Pipeline | Data validation, feature extraction, ML quality scoring, research reporting |
| AutoEIT ASR Human-Audit Pipeline | ASR plus human audit for research-grade transcription workflows |
| Robust Audio Genre Classification | PyTorch audio DL, ConvNeXt, augmentation, TTA, domain shift robustness |
| Comment Moderation NLP Ensemble | TF-IDF, sentence embeddings, gradient boosting, imbalance-aware NLP |
| LLM Code Deployment Agent | FastAPI, LLM orchestration, deployment automation, agentic workflows |
Python, SQL, TypeScript, PyTorch, scikit-learn, LightGBM, XGBoost, torchaudio, faster-whisper, pandas, NumPy, FastAPI, Docker, PostgreSQL, Vercel, GitHub Actions.
- Portfolio: jbanmol.tech
- LinkedIn: linkedin.com/in/jbanmol
- Email: jbanmol9@gmail.com