Open to opportunities

Aaradhya Mehra

@aaradhyamehra

AI/ML-focused software engineer building production RAG and LLM pipelines with verified citations and efficient caching.

India

Message

What I'm looking for

I’m looking to build production-grade ML/LLM systems—RAG, verified citations, and efficient API layers—with rigorous evaluation and load testing, working with a team that values measurable impact.

I’m a computer science undergraduate building practical, production-grade AI systems—especially RAG and LLM applications designed to reduce hallucinations and improve reliability.

In my projects, I focused on measurable outcomes. I built a support knowledge copilot with verified, hallucination-checked citations, using a hybrid retrieval pipeline (dense embeddings + BM25) and LLM-as-judge validation to boost accuracy from 72% to 88%.

I also engineered an “LLM semantic cache gateway” that detects semantically similar requests to cut cost and latency. It reduced simulated API cost by 40% and lowered P95 latency by 65%, with similarity-threshold tuning plus LLM-as-judge checks to prevent incorrect cache hits and enforce safe caching policies.

I enjoy combining strong backend engineering (FastAPI, REST APIs, Docker) with evaluation-driven development—using golden sets and load testing—so the systems I ship are not just demos, but dependable tools.

Experience

Work history, roles, and key accomplishments

Current

B.Tech in Computer Science

Current

Bangalore Institute of Technology

Sep 2023 - Present (2 years 11 months)

Undergraduate in Computer Science Engineering (CGPA 8.5/10), graduating June 2027. Builds and deploys ML/LLM models using Python, scikit-learn, and Hugging Face, with work on production-oriented AI applications using REST APIs and cloud deployment.

Python Machine Learning LLM Hugging Face scikit learn REST APIs Algorithms cloud deployment Docker fastAPI Streamlit

Augmentix Hackathon (2nd Place)

NMIT Bangalore

Jan 2026 - Present (7 months)

Secured 2nd place at the Augmentix Hackathon. Built MolGenix in 36 hours with a team, owning a FastAPI backend and an ML-based molecular docking pipeline.

fastAPI Molecular Docking Backend Development Docker

Full Stack GenAI & Agentic AI

Udemy

Completed a Full Stack Generative & Agentic AI course using Python. Covered RAG pipelines with LangChain and multi-node/stateful agents with LangGraph using vector databases.

Python Lang Chain LangGraph RAG Vector Databases

SIH Internal Qualifier (Selected)

Smart India Hackathon (SIH)

Qualified for the institute-level selection for the Smart India Hackathon by designing a federated learning solution. Led system architecture and delivered the technical presentation for the submission.

Federated Learning System Architecture Technical Presentations

C++ DSA — Abdul Bari

Udemy

Completed a C++ DSA certification focused on core data structures and algorithms. Covered graph algorithms, dynamic programming, trees, and time/space complexity analysis in C++.

C++Data Structures Dynamic Programming Algorithmic Complexity

Competitive Programming

LeetCode & Codeforces

Solved 250+ algorithmic problems across LeetCode and Codeforces, achieving a Pupil rating through rated contests.

Algorithms Problem Solving Data Structures

Machine Learning A–Z

Udemy

Completed a Machine Learning A–Z course in Python. Built regression, classification, clustering, and NLP pipelines using scikit-learn.

Python scikit learn regression Classification Clustering NLP

Education

Degrees, certifications, and relevant coursework

Bangalore Institute of Technology

Bachelor of Technology, Computer Science Engineering

2023 - 2027

Grade: CGPA: 8.5/10

Activities and societies: Projects include Support Knowledge Copilot (RAG with verified citations) and a Semantic Cache Gateway for LLM APIs; 2nd Place Augmentix Hackathon (2026), SIH internal qualifier, and 250+ LeetCode/Codeforces problems.

B.Tech in Computer Science Engineering at Bangalore Institute of Technology (CGPA: 8.5/10), graduating June 2027. Built and deployed ML/LLM-based RAG and API-based systems using Python, FastAPI, and Hugging Face.