About01Experience02Projects03Publications04Contact05
Download Resume ↓
Open to opportunities

Hello, I'm Ajaydeep

Final Year · B.Tech IT · Noida

Building production systems at the intersection of backend engineering and applied AI — from scalable APIs to end-to-end ML pipelines.

View WorkGet in Touch
Scroll
Backend SystemsAI / ML EngineeringLLM EvaluationRAG PipelinesFastAPIAWSDockerPublished ResearcherOpen to WorkBackend SystemsAI / ML EngineeringLLM EvaluationRAG PipelinesFastAPIAWSDockerPublished ResearcherOpen to Work
About

Crafting useful systems with thoughtful AI and backend engineering.

I design backend-first products and applied AI workflows that solve real problems. My work spans ML evaluation, production APIs, and scalable data systems.

Published research in a CRC Press volume
Built clinical ML pipelines with reproducibility in mind
Delivered real-time data and aggregator systems

My Tech Stack

TS

TypeScript

React

Next.js

FastAPI

Node.js

REST APIs

PostgreSQL

MongoDB

Redis

Docker

AWS

CI/CD

GitHub

PyTorch

TensorFlow

ETL Pipelines

Data Modeling

HTML/CSS

Vercel

02 — Experience

Where I've worked.

June 2025 — July 2025

Internship · Noida

Machine Learning Intern

@ CoE-AIE
  • Built an end-to-end multimodal ML pipeline using audio and textual data from 140+ real-world clinical interview sessions.
  • Automated speech isolation and feature extraction workflows, reducing preprocessing time by ~40%.
  • Engineered and evaluated paralinguistic and textual features using COVAREP, training models with improved F1-score.
  • Designed reproducible training and evaluation workflows with feature-importance–driven error analysis.
PythonCOVAREPFeature EngineeringNLPScikit-learn

Nov 2024 — Apr 2025

Freelance

LLM Evaluation Analyst

@ Outlier
  • Evaluated LLM outputs for hallucination, factual accuracy, reasoning and policy adherence across diverse task types.
  • Designed adversarial and edge-case prompts to stress-test LLM behavior and identify systematic failure modes.
  • Applied prompt engineering techniques (few-shot, CoT) to improve output consistency across task domains.
  • Conducted qualitative and quantitative analysis of response quality to improve LLM evaluation workflows.
LLM EvaluationPrompt EngineeringAdversarial TestingCoTRLHF

03 — Projects

Selected work.

Click to expand ↓

01

Orderly

Full Stack · AI+

Overview

A real-time e-commerce aggregator unifying product search across Amazon, Flipkart, Blinkit, and Zepto. Built a retrieval and matching pipeline using TF-IDF, attribute filtering, and fuzzy similarity achieving 85% match accuracy. The backend handles concurrent requests with sub-second latency via Redis caching and MongoDB indexing.

Period

Jan 2025 — May 2025

Highlights

  • TF-IDF + fuzzy similarity pipeline with 85% match accuracy
  • Sub-second latency under concurrent load via Redis caching
  • FastAPI backend deployed on AWS with Docker containers
  • Presented at IC3-2025 International Conference

Stack

PythonFastAPIPlaywrightMongoDBRedisDockerAWSVercel
02

SageML

ML · Research+

Overview

A modular, end-to-end ML orchestration framework built for automated training, evaluation, and GenAI-oriented workflows. Designed with RAG- and LLM-oriented abstractions supporting both structured and unstructured data with configurable NLP pipelines. Published as a book chapter in CRC Press's Generative AI: Technology and Applications.

Period

Aug 2024 — Nov 2024

Highlights

  • Modular pipeline abstractions for structured + unstructured data
  • RAG and LLM-oriented design with grounding and evaluation focus
  • Configurable NLP pipelines with automated preprocessing
  • Published in CRC Press book chapter (Chapter 8)

Stack

PythonPyQt6PandasNumPyScikit-learnTesseract OCR

04 — Publications & Achievements

Research & recognition.

Published Work

Book Chapter2024

SageML: An Automated Machine Learning Pipeline Framework

Generative Artificial Intelligence: Technology and Applications, Chapter 8

CRC Press

Conference Paper2025

ORDERLY: An Intelligent Hybrid Platform for Online Shopping System

17th International Conference on Contemporary Computing (IC3-2025)

Noida, India

Certifications & Achievements

Data Science Professional 2025

Oracle Certified Professional

Principles of Generative AI

Infosys Springboard

API Fundamentals Student Expert

Postman Academy

State Rank 8 — National Science Olympiad

Achievement

05 — Contact

Let's build
something.

Open to internships, full-time roles, and interesting collaborations. Whether it's a backend system, an AI pipeline, or a research project — reach out.