I'm an ML engineer at DTU (Mathematics & Computing) who builds systems that work under real-world constraints — competitive hackathon environments, bioinformatics pipelines, and financial forecasting with production-grade anomaly monitoring. I care as much about latency and compute cost as I do about model accuracy.
Hackathon record: Won Brainwave 2.0 MLH (200+ competitors) · Top 1% at Smart India Hackathon 2025 (1,000+ participants)
Every commit I push is GPG-signed and DCO-compliant. This isn't optional hygiene for me — it's how I've worked since day one.
Open to remote ML/AI engineering roles and serious freelance collaborations. If your team cares about reproducibility and production correctness, we'll get along.
|
Real-Time Deepfake Detection
Won Brainwave 2.0 MLH Hackathon (Jan. 2026, 200+ competitors). Agentic router dispatches media across 4 specialist models via LangGraph. Reduced compute cost 3× via knowledge distillation into MobileNetV3. 95%+ accuracy · <3s latency View on GitHub → |
Self-Supervised DNA Discovery
Top 1% at Smart India Hackathon 2025 (1,000+ participants). Classified 10,000+ unlabeled DNA sequences using a self-supervised Transformer, eliminating manual annotation entirely. Cut graph latency by 30% and memory overhead by 25% via approximate nearest-neighbor algorithms. View on GitHub → |
AI Financial Forecasting Platform
Multi-model forecasting ensemble (Holt-Winters, OLS, Seasonal-Naive) achieving 90% prediction coverage on SME and agricultural banking datasets. Integrated rolling Z-score anomaly monitoring, regime-shift tracking, and Gemini 2.0 for plain-English business impact summaries. View on GitHub → |
|
Fine-tuned CNN on Food-101 (101 categories) with production inference deployed to Vercel.
|
NLP pipeline using TF-IDF + Logistic Regression with Streamlit interface for real-time classification.
|
|
Core ML/AI
Data & Analysis
|
Infrastructure & Deployment
Verified workflow
|
- Deeper work on agentic AI architectures — multi-model routing, tool-use, and evaluation pipelines
- Building a personal MLOps scaffold with experiment tracking, model versioning, and automated regression testing
- Contributing to open-source with strict reproducibility and correctness standards
I work in environments that enforce strict provenance standards. Every commit in my repositories is:
- ✅ GPG-signed — cryptographic proof that the commit came from me, verifiable by anyone
- ✅ DCO sign-off included —
Signed-off-byon every commit, certifying origin under the Developer Certificate of Origin
If you're a maintainer reviewing a contribution from me, you will never chase me for these.
B.Tech — Mathematics and Computing (MCE)
Delhi Technological University (DTU) · Expected 2028
A dual-focus program that grounds ML work in rigorous linear algebra, probability, and optimization theory — not just API calls.