B.Tech AI & DS — IIT Jodhpur

Jaineel
Vora

Building intelligent systems, from distributed infrastructure to machine learning.
Interested in backend engineering, AI systems, and the layers that connect them.

8.72+
CGPA
1st
Inter IIT Tech Meet 14.0
250+
Users — libr beta
Scroll to explore
01

About

I'm a third-year B.Tech student in Artificial Intelligence and Data Science at IIT Jodhpur.

My interests lie at the intersection of backend systems and machine learning. I've built distributed applications using peer-to-peer protocols, engineered graph-powered platforms, fine-tuned language models, and designed AI systems that operate under real-world constraints. I'm drawn to problems that require both strong systems thinking and practical ML.

I was part of the team that won 1st place at Inter IIT Tech Meet 14.0 for designing a client-centric WiFi resource management framework combining edge inference, reinforcement learning, and network optimization.

Outside of coursework and competitions, I enjoy building, experimenting, and learning by doing. I'm fascinated by how systems scale, evolve, and remain reliable as they grow in complexity.

If you're working on something interesting or just want to talk about distributed systems, backend engineering, or AI, I'd love to connect.

02

Technical Expertise

Education
B.Tech — AI & Data Science
Indian Institute of Technology, Jodhpur
CGPA 8.72 2024 – Present
Senior Secondary (HSC)
Smt. Shanti Devi Shukla Junior College of Science
90.50% 2024
Secondary (ICSE)
Kapol Vidyanidhi International School
98.40% 2022
Skills
Languages
PythonC/C++GoJavaScript
ML & AI
PyTorchQLoRADPO/PPOVLMRAG
Backend & DB
DjangoFastAPIPostgreSQLNeo4jPineconeRedis
Systems & Infra
libp2pKademlia DHTWebSocketsDockerNAT
Relevant Coursework
Data Structures & Algorithms Operating Systems Computer Networks Computer Architecture Pattern Recognition & Machine Learning Probability & Statistics Linear Algebra Introduction to Computer Science
03

Experience

May 2026 — Present
Research
Undergraduate Researcher
Guide: Prof. Angshuman Paul · IIT Jodhpur
  • Training a diagnostic Vision-Language Model (VLM) on a custom multimodal dataset aligning medical imaging with noisy, multilingual Indic layperson dialogue.
  • Focused on mitigating omniscient patient data leakage and correcting linguistic biases introduced by multilingual inputs.
Oct 2025 — Dec 2025
Competition
Project Member — Inter IIT Tech Meet 14.0 1st Place
Arista Networks Problem Statement
  • Co-designed a client-centric Wi-Fi RRM system using graph-based optimizations and safe reinforcement learning across 3 control loops — 98.7% safety compliance, 4.4h adaptation cycle.
  • Built an explainability chatbot and CNN interference classifier with 97% accuracy and <14ms latency.
  • Benchmarked on 3 heterogeneous multi-AP pilot sites via NS-3 and MATLAB: +28% edge throughput, -40% P95 latency, +16.3% RU utilization, -30% hidden-node collisions.
04

Achievements

🥇
1st Place — Inter IIT Tech Meet 14.0
Won Gold among 23 IITs for the Arista Networks problem statement, delivering a full Wi-Fi RRM system with explainable AI and rigorous simulation benchmarks.
🥈
2nd Place — Kaggle OTAKU AI DŌJŌ
Fine-tuned Vision Transformers (ViT) on 93K+ images, filtering out-of-domain outliers to handle domain shifts across 630 classes and 6 unseen styles.
🥉
3rd Place — Insomniac 24-Hour Hackathon
Engineered BEACON, an AI mentorship platform with async RAG pipelines, Pinecone vector search, and 2-hop Neo4j graph traversal.
Core Member — Devlup Labs, IIT Jodhpur
Contributing to IIT Jodhpur's developer community, shipping open-source projects, and participating in inter-IIT collaborative builds.
05

Projects

01 · Mar 2026 – Apr 2026
BEACON
AI-Driven Mentorship Platform · Insomniac Hackathon (3rd Place)
Full-stack AI mentorship platform with Google OAuth/JWT, real-time WebSockets sustaining 200+ concurrent connections, and Neo4j-based mentor discovery achieving 92%+ match acceptance via trust rankings. Async RAG pipeline with 95% Top-3 retrieval accuracy and ~65% lower recommendation latency.
PythonDjangoReactPostgreSQLNeo4jRedisCeleryPineconeGeminiWebSockets
02 · Oct 2025 – Nov 2025
GraphAnalyse
Social Graph Engine · DSA Coursework Project
Scalable, concurrent in-memory graph engine — <1ms latency for 100K+ nodes using Bipartite PageRank, Jaccard similarity, and Min-Heaps. Reduced memory footprint by 92% via HyperLogLog and Inverted Indices. 2M queries/sec autocomplete via Tries, 163M ops/sec via DSU.
C++React.jsGraph AlgorithmsRaw SocketsHyperLogLogPageRank
03 · Jun 2025 – Oct 2025
libr
Decentralized Public Forum · Summer of Code — Devlup Labs
Custom P2P protocol via Kademlia DHT (XOR routing) with <50MB memory and <100ms discovery latency. Adaptive relay fallbacks achieving 95%+ NAT-traversal success. Shipped a beta release onboarding 250+ users and generating 1,000+ posts across multiple operating systems.
Golibp2pKademlia DHTP2P Networks
04 · Jan 2025 – Apr 2025
NL → MongoDB
NLP Query Engine · Winter Project — AI Club × MetaFusion
Curated 1,000 complex query pairs across 10 multi-collection schemas using GPT-4o Mini and a validation agent. Fine-tuned to 94% accuracy via 4-bit QLoRA (Unsloth+TRL) with DPO/PPO preference alignment. Optimized for 16GB CPU with llama.cpp — inference latency under 500ms.
Fine-TuningNLPQLoRADPOPPOPrompt Cachingllama.cpp
06

Blogs

Building a Client-Centric WiFi Resource Management System for Enterprise Networks

How we designed a client-centric WiFi RRM framework using dedicated spectrum sensing, edge intelligence, cloud-scale optimization, and machine learning.

Read Blog

From Natural Language to MongoDB: Building a Fast, Private Query Engine for Real-World Use

Every database project has the same hidden friction point: the data may be powerful, but the interface is not always human-friendly.

Read Blog

Winning 2nd Place in OTAKU AI DŌJŌ: Building a Robust Anime Character Classifier Under Style Shift

OTAKU AI DŌJŌ was one of those competitions that looked straightforward at first glance and then quickly turned into a problem about everything except “just classification.”

Read Blog
07

Contact

Let's
Build
Something.

Open to research collaborations, internships, and interesting conversations about systems, AI, or anything you're working on.

Contact Me
Send a Message