Resume - Ayush Chakraborty

Education

Bachelor of Technology in Computer Science and Engineering Expected May 2027

PES University, Bangalore

CGPA: 9.28 / 10.0 (Top 5% of batch)

Relevant Coursework: Machine Learning, Database Management System, Big Data, Software Engineering, Computer Networks, Operating Systems

Research Intern — The Innovation Lab, PES University June 2025 – August 2025

Selected as 1 of 26 interns from 1000+ applicants.

Co-developed SUBGNN with 3 research interns: a scalable subgraph isomorphism framework that accelerates search on large-scale graphs via a filter-and-verify architecture.
Designed a multi-stage pipeline using METIS for graph partitioning and Graph Neural Networks (GNNs) to generate topology-aware vector embeddings for efficient candidate filtering.
Built the retrieval stage using FAISS to prune the search space, achieving an 82.5% mean speedup (on CORA) and 91.18% node mapping accuracy (on OGBN-Arxiv, 170k nodes) compared to traditional solvers.

Transformer Model & C Serving Layer — C, PyTorch, Python Dec 2024 – March 2025

Implemented a decoder-only Transformer architecture from first principles in PyTorch, incorporating multi-head self-attention and positional embeddings to perform autoregressive language modeling.
Built a C serving layer to execute the Python inference process via standard I/O streams and achieved real-time token streaming by manipulating standard I/O buffers to pipe generated tokens immediately to the client, replicating the user experience of production LLMs.

CRDT-Based Distributed Key-Value Store — Rust, CRDTs, Tokio Sep 2025 – Present

Co-developed a leaderless, eventually consistent key-value store in Rust, leveraging Conflict-free Replicated Data Types (CRDTs) to enable mathematically verifiable convergence without central coordination.
Built an asynchronous P2P gossip protocol for highly concurrent node-to-node state synchronization and anti-entropy mechanisms.

Redis-like In-Memory Store — C++, TCP Sockets, Event Loop Oct 2025 – Present

Built a single-threaded, event-driven server using non-blocking I/O to handle concurrent clients without context-switching overhead.
Implemented a custom chained hash table for O(1) lookups and the RESP serialization protocol to manage efficient binary data transmission.

Distributed Systems Implementation (MIT 6.824 Labs) — Go, RPC, Concurrency Nov 2025 – Jan 2026

Built a fault-tolerant MapReduce system handling worker failures and stragglers via a custom RPC coordinator.
Developed a Raft consensus module to guarantee strong consistency across sharded key-value pairs, implementing leader election and log replication.

Languages: C++, C, Rust, Python, Go

Technologies: PyTorch, Docker, Tokio, Redis, FAISS, Git, Linux/Bash, MySQL, MongoDB, LaTeX

Certifications: Deep Learning Specialization, Machine Learning Specialization (DeepLearning.AI)

Prof. MRD Merit Scholarship: Awarded consecutively (Sem 1–4) for maintaining rank in the top 5% of the batch.
State Merit Scholar: Ranked 1437 out of 200,000+ candidates (Top 0.7%) in the State Engineering Entrance Exam (KCET).