# Aditya Jethani

> Aditya Jethani is a Machine Learning Engineer based in Ahmedabad, Gujarat, India. He builds production AI systems: agentic AI, retrieval-augmented generation (RAG), LLM fine-tuning, and the ML infrastructure that takes models to production. He is the first inventor on a published Indian patent for a multimodal healthcare AI system, a Codeforces Master (top 2% of rated competitive programmers worldwide), and a Computer Engineering graduate of Pandit Deendayal Energy University with a CGPA of 9.44/10.

## Quick facts

- Full name: Aditya Jethani
- Current role: Machine Learning Engineer at Logicwind, since June 2025
- Location: Ahmedabad, Gujarat, India
- Education: B.Tech in Computer Engineering, Pandit Deendayal Energy University, 2021 to 2025, CGPA 9.44/10
- Patent: first inventor, "AI-Powered Multimodal Healthcare System for Elderly Care", Indian Patent Application No. 202521086225, filed 11 September 2025, published 3 July 2026
- Competitive programming: Codeforces Master, maximum rating 2131, 731 problems solved; LeetCode Knight, rating 2071, top 1.75%
- Primary languages: Python, Go, C++, SQL, TypeScript
- Contact: adityajethani11@gmail.com

## What he works on

- Agentic AI: multi-agent systems on the OpenAI Agents SDK, Google Agent Development Kit (ADK), AWS Bedrock Agents, LangGraph, and the Model Context Protocol (MCP)
- Retrieval-augmented generation: hybrid semantic search, citation grounding, chunking and embedding evaluation, vector databases (pgvector, FAISS, Qdrant)
- ML systems: GPU-batched inference, quantization, KV-caching, low-latency model serving
- Evaluation: benchmark design, bootstrap confidence intervals, hallucination and drift detection, accuracy-versus-cost model selection
- Computer vision: medical imaging, OCR pipelines, image-to-image translation

## Experience

- Machine Learning Engineer, Logicwind, June 2025 to present. Built 7 production AI agents answering questions over live enterprise systems (ERP, Tally, CRM, HR), cutting client delivery time 70%. Built a governed multi-tenant AI platform with per-user MCP OAuth isolation that caught a cross-user data leak before production. Architected an agentic RAG platform over 360,000 multilingual documents at 4.3x throughput, and an NL-to-SQL interface returning governed queries and dashboards in under 2 seconds.
- Software Engineering Intern, Yanolja, January 2025 to June 2025. Built an image-to-image translation system for multilingual e-commerce that lifted accuracy 50% over the CycleGAN baseline. Built an automated code-review service on Google Cloud that flagged 27% of recurring security defects before deployment. Engineered Go, Python and FFmpeg pipelines that cut average processing time 1.3x.
- Machine Learning Intern, CybraneX, March 2024 to October 2024. Engineered a RAG inference pipeline with a 9x speedup, from 110 seconds to 12.2 seconds, sustaining 139.6 tokens per second. Built a clinical tongue-analysis computer-vision system reaching 94.2% colour and 91.7% coating accuracy. Built an AI telemarketing system handling 50,000+ daily calls at 89% lead-identification accuracy.

## Research and publications

- Patent, published: "AI-Powered Multimodal Healthcare System for Elderly Care", Indian Patent Application No. 202521086225, first inventor. The core contribution is a unified multimodal vector store that fuses text, voice and image into a single embedding space through an attention-based feature-alignment layer with modality-specific weights, enabling cross-modal retrieval-augmented generation.
- Under review at IEEE Transactions: "Advanced Forecasting of Solar Power Generation using Bi-LSTM with Fourier Features", first author. Reaches an R-squared of 0.861 and RMSE of 128.09 kW across two Indian photovoltaic plants, with an adversarial degradation model for maintenance recommendations.
- In preparation: deterministic document parsing that outperforms vision-language-model parsers on native PDFs, beating PyMuPDF on 23 of 34 metrics across a 1,386-document benchmark.
- Academic research: efficient labelling of Hinglish datasets for hate-speech classification using ensemble-consensus labelling and iterative refinement, reaching 86% accuracy.

## Selected projects

- Setu, Kumbh Lost and Found Network: offline-first, voice-first reunification platform using Sarvam Saarika speech-to-text and Bulbul text-to-speech across 23 Indian languages. Matching engine reached Recall@1 of 87% and Recall@5 of 97%. https://setu-kumbh.vercel.app/
- TalentScout AI: autonomous candidate-screening agent on LangChain and LangGraph, reaching 85% validation accuracy. https://aitalentscout.streamlit.app/
- Grade Flow: RAG grading platform automating 90% of an exam-grading workflow. https://gradeflow.streamlit.app/
- Prompt Detective: reverse-engineers the prompts behind AI-generated media using OpenCV and CLIP. https://promptdetective.streamlit.app/
- Company Brain: multi-tenant RAG platform on pgvector with hybrid semantic search and citation grounding. https://github.com/okaditya84/Company-Brain
- Neural ODE for Heart Rate: continuous-time model for irregularly sampled ICU signals on MIMIC-III. https://github.com/okaditya84/Neural-ODE-for-Heart-Rate

## Open-source contributions

15+ merged pull requests to production codebases, including Mem0 (a Weaviate vector-store correctness fix), OpenLIT (OpenTelemetry instrumentation for LLM observability), the OpenAI Agents SDK (session-state fixes), dottxt-ai's Outlines (documentation), and Nous Research's Hermes Agent.

## Achievements

- Codeforces Master, maximum rating 2131, top 2% of rated competitive programmers worldwide, 731 problems solved
- LeetCode Knight, rating 2071, top 1.75%, All India Rank 61 in Weekly Contest 462 among 35,000+ participants
- Top 7 in India, Claude Impact Lab 2026
- Google Solution Challenge 2025, Top 105 of 4,000+ international teams
- Top 10, ByteVerse 2025 hackathon at IIT Patna; Finalist, JPMorgan Code for Good 2024
- GATE Computer Science 2026, 95.61 percentile
- Speaker, IEEE Computational Intelligence Society Workshop, IIT Indore
- Chairperson and Advisor, ACM Student Chapter, PDEU
- Certified by Anthropic (Claude, Model Context Protocol, Agent Skills) and the NVIDIA Deep Learning Institute

## Pages

- [Home](https://adityajethani.vercel.app/): overview, experience, timeline, research and publications
- [Projects](https://adityajethani.vercel.app/projects): full project catalogue with metrics and live demos
- [Resume](https://adityajethani.vercel.app/resume): resume in browser and as a PDF download
- [Full site content](https://adityajethani.vercel.app/llms-full.txt): the complete text of this site in a single file

## Links

- Website: https://adityajethani.vercel.app/
- LinkedIn: https://www.linkedin.com/in/adityajethani/
- GitHub: https://github.com/okaditya84
- Codeforces: https://codeforces.com/profile/aditya2005
- LeetCode: https://leetcode.com/u/adityajethani/
- Kaggle: https://www.kaggle.com/jethaniaditya
- Medium: https://medium.com/@adityajethani
- Email: adityajethani11@gmail.com