I build machine learning systems where performance is a first-class citizen. My work sits at the intersection of Systems Programming (Rust/C++) and Applied ML Research, with a specific focus on Concept Drift Adaptation and Hardware-Aware Optimization.

I prefer building quietly, measuring carefully, and optimizing relentlessly.

I specialize in MLSys—designing the computational engines that power modern AI:

• Algorithmic Research: Concept Drift Adaptation and Self-Adaptive Gradient Boosting for non-stationary data streams
• Low-Level Systems: High-performance kernels using SIMD, multi-threading via Rayon, and memory-safe systems in Rust
• Optimization: Numerical optimization (Newton-Raphson), information theory (Shannon Entropy), and sub-millisecond similarity search
• Infrastructure: Asynchronous, high-throughput backends using Axum and Tokio

Self-Adaptive Gradient Boosting Library in Rust | Mentored by Ash Vardanian (Founder, Unum Cloud)

PKBoost is a production-ready GBDT framework built to handle "drifting" data where standard models fail.

Technical Innovations:

• Shannon entropy-guided splitting with second-order Newton optimization
• Metamorphic adaptation: real-time tree vulnerability tracking and selective pruning/retraining
• SIMD-accelerated kernels (SimSIMD) with Rayon parallelism
• PyO3 bindings for seamless Python integration

Empirical Results:

• Only 2.8% PR-AUC degradation under severe concept drift vs. 12-18% for XGBoost/LightGBM
• 2,400+ PyPI downloads and permanently archived (DOI: 10.5281/zenodo.17568991)

Links: GitHub | PyPI | Preprint

Built a production-grade RAG backend for HR automation (demoed to Godrej Living):

• Architecture: High-throughput async backend using Rust, Axum, and Tokio
• Vector Search: Custom in-memory vector store with USearch (HNSW) and FastEmbed for sub-millisecond search
• Agentic Logic: ReAct-style agent loop with Llama 3.3 for dynamic tool dispatching
• Persistence: PostgreSQL with SQLx for ACID-compliant chat history and vector storage

Enterprise B2B prioritization system deployed at Value Score:

• Performance: 0.89 ROC-AUC with 1-2 minute training time on 10k leads
• Efficiency: <2GB RAM footprint with native categorical handling
• Real-World Impact: 12x efficiency improvement (5% → 60% conversion rate), 80% sales time saved
• Tech Stack: CatBoost, Zoho CRM, automated feature engineering pipeline

Value Score Business Solutions LLP | Technical Intern | Jun 2025 – Present

• Building agentic RAG workflows with n8n and open-source LLMs
• Developed custom Rust RAG agent for HR automation
• Evaluated Zoho ecosystem for AI/ML production deployment

Artech Communications | Network Engineering Trainee | Dec 2024 – Apr 2025

• Configured high-availability hospital network infrastructure
• Administered servers and security testing for critical systems

Outside of code, I am an Amateur MMA District Gold Medalist with a 5-1 record. Engineering and combat sports share the same DNA:

• Pressure Testing: A system's reliability is only proven when stressed to its limits
• Fundamental Mastery: Deep knowledge of data structures, operating systems, and mathematics over framework hype
• Relentless Improvement: Building elite systems requires daily discipline

Other Achievements:

• Amateur MMA District Gold Medalist, Mumbai 2022 (5-1 record)
• TRCAC Chess Gold Medalist 2024

I'm open to ML Systems Engineer, Applied AI Researcher, or Performance Engineering roles where technical discipline is the standard.

BUY ME A COFFEE : Buy me a coffee "Build quietly. Measure carefully. Improve relentlessly."