---

Five years shipping AI from research to production.

Currently building [AURA] — a 30B multimodal foundation model generating 2K video + synchronized stereo audio in a single forward pass, with sampling cut from 50 steps to ≤8 via Rectified Flow + step distillation.

"A model that only works at fifty sampling steps is a research result, not a product."

The through-line across all of it: inference budgets — getting capable models to run fast, cheap, and reliably outside the lab.

---

🔭 What I'm Building

🎬 AURA 30B multimodal foundation model. 2K video + stereo audio in one forward pass. 50-step sampling distilled to ≤8 via Rectified Flow + Latent Consistency Models. PyTorch DiT Rectified Flow Distributed Training

📜 Sigil Mobile app converting spoken agreements into binding micro-contracts. The LLM runs entirely on-device — zero cloud round-trips, complete privacy. On-device SLMs Speech Processing Privacy-first

⚡ SemantiCache Semantic LLM caching. 61% cache hit rate on real chatbot data. Cut API costs 40–70% with one Python decorator. Zero external dependencies. Python Sentence Transformers SQLite MIT

---

🌍 Open Source

Correctness fixes over cosmetic changes — every fix ships with a failing-then-passing test:

Repository	Stars	Contribution	Status
feast	7k+ ⭐	Redis event-time ordering · pymssql arm64 wheels · single-lookup feature-view resolve · RemoteRegistry client-side cache	✅ 2 merged · 🔄 2 in review
statsmodels	11k+ ⭐	SARIMAX extend() specification-validation fix	✅ Merged
recharts	27k+ ⭐	Rectangle kept mounted while it animates to zero	🔄 In review
Gymnasium	12k+ ⭐	Text space multi-character charset validation	🔄 In review
folium	7k+ ⭐	Glyphicons-only stylesheet (drop full Bootstrap) · review & triage	🔄 In review

---

🧠 Tech Stack

Generative AI & ML

Languages & Frameworks

Cloud & MLOps

---

📊 Stats

---

📌 Featured

---

_{📍 Bengaluru, India  ·  Open to collaboration on inference optimization, multimodal generation, and impactful open source}