Ashish Patil

I am an Electronics and Communication Engineering undergraduate at Visvesvaraya National Institute of Technology (VNIT), Nagpur, with a strong foundation in mathematical modeling and quantitative analysis.

Driven by an entrepreneurial spirit, I successfully secured an institutional grant (60,000 INR) from VNIT to build an independent technical MVP—reducing campus attendance processing times by 98% through hardware IoT and real-time facial recognition.

My core focus lies at the intersection of statistical methods, algorithmic problem-solving, deep learning architectures, and data assimilation, with a deep interest in applying quantitative engineering to complex system modeling and capital markets.

1. June 2025 – June 2026

   Co-Founder / Technical Lead · Nexalaya (Campus Digitization MVP)

   Nagpur, Maharashtra

   • •Conceptualized and pitched a campus digitalization initiative, successfully securing a 60,000 INR grant from VNIT to build the Minimum Viable Product (MVP).
   • •Designed a scalable system architecture integrating hardware IoT modules with real-time facial recognition, delivering a 98% reduction in attendance processing time (from 10 minutes to 10 seconds per class).

   • System Architecture
   • IoT Hardware Modules
   • Facial Recognition
   • Grant Funding
   • Python
   • C++
2. Mar 2026 – Apr 2026

   Web & Business Development Intern · Bookstaken

   Nagpur, Maharashtra

   • •Executed 200+ strategic sales calls, analyzing client feedback to adjust outreach strategies, resulting in a 20% boost in lead generation.
   • •Developed backend infrastructure and interactive modules for the core platform, optimizing database performance to increase traffic retention.

   • Backend Infrastructure
   • Database Optimization
   • Lead Generation
   • Sales Strategy
   • JavaScript
   • Node.js
3. Mar 2026 – Apr 2026

   Backend Development Trainee · Skillified Mentor

   Remote

   • •Architected secure, scalable data pipelines and authentication APIs, ensuring robust data handling and system stability under load.
   • •Built a blog API supporting CRUD operations, pagination, search and authentication for protected routes, gaining a thorough understanding of database schemas and search efficiency.

   • Data Pipelines
   • Authentication APIs
   • CRUD Operations
   • Pagination & Search
   • Database Schemas
   • System Architecture

• Chatly Platform — Full-Stack Realtime Messaging Engine

  Production-grade realtime chat application delivering sub-100ms message delivery, optimistic UI with server reconciliation, JWT auth rotation, and scalable pub/sub architecture.

  ⚡Delivery Latency:<100ms

  🐳Docker Services:4 Nodes

  🔄Auto Data Reset:24h Cycle

  • •Engineered sub-100ms realtime messaging infrastructure using Socket.IO, WebSocket transport, and @socket.io/redis-adapter for multi-node horizontal scaling.
  • •Architected optimistic UI state reconciliation in React 18 & Zustand for instant message delivery, threaded replies, emoji reactions, and soft-deletes.
  • •Implemented secure auth with 15-min RS256 JWT access tokens, HTTPOnly refresh token rotation backed by Redis blacklisting, and Zod runtime schema validation.
  • •Built S3-compatible pre-signed upload pipelines with Supabase Storage, Prisma ORM connection pooling, cursor-based pagination, and full Docker Compose orchestration.

  Datasets:

  Supabase PostgresRedis 7Socket.IO 4.7Docker

  Live Demo

  • TypeScript
  • React 18
  • Socket.IO
  • Node.js
  • Express
  • Supabase Postgres
  • Redis
  • Prisma ORM
  • Zustand
  • Tailwind CSS
  • Docker
  • Zod
• Predictive Modeling & Optimization: Image Super-Resolution with FSRCNN

  High-efficiency deep learning hourglass architecture engineered for real-time super-resolution without external bicubic pre-processing.

  ⚡Complexity Reduction:98%

  📊Benchmark Datasets:4 Sets

  • •Engineered a mathematical FSRCNN model in PyTorch utilizing an hourglass architecture—incorporating feature extraction, 1x1 convolutional shrinking for dimensionality reduction, and deconvolutional upsampling.
  • •Optimized model accuracy by calculating a weighted combination of pixel-wise Mean Squared Error (MSE) and VGG-16/19 Perceptual Loss to evaluate complex spatial features.
  • •Systematically reduced computational complexity and memory consumption against baseline SRCNN models by eliminating external bicubic interpolation and learning upscaling operations internally.
  • •Developed custom data extraction pipelines for large-scale datasets (BSD-100, Set-5, Set-14, Urban100) and maximized parameter efficiency to enable flexible scalability across multiple scaling factors.

  Datasets:

  BSD-100Set-5Set-14Urban100

  Source Code

  • PyTorch
  • Convolutional Neural Networks
  • OpenCV
  • Perceptual Loss
  • Mathematical Modeling
• High-Frequency Limit Order Book & Ultra-Fast Matching Engine

  A production-grade, deterministic, low-latency Limit Order Book (LOB) and Matching Engine engineered for algorithmic trading venues adhering to Price-Time Priority (FIFO) matching rules.

  ⚡Peak Throughput:900k+ ops/sec

  ⏱️Mean Latency:781 ns

  💾Hot Path Allocations:0

  • •Architected an in-memory Limit Order Book and Matching Engine in C++17 capable of processing 900,000+ simulated order operations per second.
  • •Achieved 781 ns mean matching latency and eliminated dynamic heap allocations on the hot path via custom intrusive slab pools.
  • •Engineered a zero-allocation fixed-size slab memory allocator that pre-allocates a contiguous memory slab of 65,536 Order structures at initialization.
  • •Implemented intrusive doubly-linked price levels for instantaneous O(1) order enqueue operations and O(1) order cancellations.
  • •Enforced mathematical determinism by utilizing fixed-point scaled integer pricing, multiplying all prices by 100 to eliminate floating-point imprecision.
  • •Built an interactive real-time React trading dashboard featuring a live L2 depth ladder, trade tape, and automated simulation bot via a native HTTP REST API.

  Datasets:

  100,000 Randomized Order Operations Benchmark

  Live Dashboard

  • C++17
  • React
  • Vite
  • Docker
  • Native HTTP/1.1 Socket Server

1. 2025

   VNIT Institutional MVP Grant

   Visvesvaraya National Institute of Technology, Nagpur

   🏆 60,000 INR Institutional Funding

   Secured competitive grant funding from VNIT Nagpur after pitching a campus digitalization initiative, designing an IoT and facial recognition architecture that cut attendance time by 98%.

2. Ongoing

   Algorithmic & Quantitative Problem Solving

   LeetCode & Competitive Coding

   🏆 Active Competitive Profile

   Consistently solves complex algorithmic challenges in data structures, dynamic programming, mathematical modeling, and graph theory.