ISHAAN
SHAILESH PHAYE
Engineering production-ready AI systems, custom database storage engines, and high-concurrency distributed backends. Specializing in autonomous RAG platforms, fine-tuned LLM architectures, and systems programming.
Production-grade RAG pipelines, fine-tuned transformer architectures, cross-encoder re-ranking, and low-latency edge computer vision.
Custom storage engines, disk I/O, cache-conscious page layouts, concurrency, and rigorous algorithmic problem solving (400+ LeetCode).
Asynchronous REST APIs, responsive reactive frontends, containerized microservices, and relational/document databases.
Machine Learning Intern @ Datanomers
Engineered high-scale Python data preprocessing pipelines, optimized real-time predictive ML inference algorithms, and collaborated on banking-domain models.
M.Tech in Computer Science
Indian Institute of Information Technology, Lucknow (IIIT Lucknow)
B.Tech in Information Technology
Shri Shankaracharya Institute of Professional Management & Technology (SSIPMT)
Engineered a production-ready RAG platform transforming YouTube videos and playlists into queryable knowledge bases. Features a two-stage retrieval pipeline (Qdrant dense vector search + cross-encoder re-ranking) operating within a sub-80 MB serverless RAM footprint. Implemented concurrent thread-pool retrieval across 35+ playlists, cutting search latency by 96% (8s down to <0.3s) with dual-token JWT auth and timestamped citations synced to an embedded player.
Engineered the complete GPT-2 (124M/355M) architecture from scratch in PyTorch covering causal multi-head attention, Byte Pair Encoding tokenization, and pre-LayerNorm residual blocks, transferring OpenAI pretrained weights into the custom model. Fine-tuned checkpoints for sequence classification (95.67% test accuracy) and instruction following via SFT on the Alpaca dataset; built an autoregressive inference engine with Top-k sampling and temperature control.
Developed a responsive full-stack e-commerce web platform for a regional startup from structural ideation to production deployment, owning the entire SDLC. Designed schema architectures and compound indexing strategies in MongoDB, reducing backend API database query processing latency by 20% while containerizing services with Docker.
Engineered a disk-backed key-value store in C++ from scratch featuring a custom B+Tree index, supporting O(log n) point lookups and range scans across leaf-node linked lists with cache-conscious page layouts. Built Write-Ahead Logging (WAL) with sequential log flushing prior to in-place page writes, guaranteeing ACID atomicity, crash-safe durability, and automated recovery on restart alongside an LRU buffer pool manager.
Led development of an on-device computer vision pipeline detecting and classifying objects with 93% accuracy, deployed as a real-time mobile application. Trained an SSD MobileNet model on the MSCOCO dataset and exported quantized models to TFLite format for edge deployment, achieving ultra low-latency inference directly on hardware without cloud dependencies.
8th Place / 3,500+ Participants
Secured 8th position among 3,500+ developers at TechXR Hackathon (IIT Roorkee), earning a competitive 3-month internship offer based on technical architecture and performance.
400+ Algorithmic Problems Solved
Solved over 400 algorithmic problems on LeetCode focusing on asymptotic runtime and memory complexity optimization across Dynamic Programming, Graphs, and Trees.
VIEW PROFILE @ishaanroxx ↗Top 200 All-India Rank
Attained a Top 200 national percentile rank in TCS National Qualifier Test (NQT) nationwide with only one month of targeted algorithmic preparation.
Actively seeking roles across Software Development Engineering (SDE), Backend Systems, and AI / Machine Learning Engineering.