Open to AI/ML and software engineering roles
Building AI that ships, not just demos.
Recent MSc AI graduate (Distinction). I design and deploy production-ready AI services: custom orchestration engines, RAG pipelines and fine-tuned deep learning models in Python.
- 81%
- Dissertation, High Distinction
- 75%
- MSc overall, Distinction
- 3
- Open-source LLMs fine-tuned
- 23×
- Accuracy improvement, 1.3B model
From neural network to production API.
A highly driven AI specialist with hands-on experience building custom orchestration engines, designing RAG pipelines and fine-tuning deep learning models in Python. I want to solve real-world problems inside a high-performance engineering team.
- Agentic AI
- Tool calling and iterative search loops
- RAG
- ChromaDB over local documents
- LLM orchestration
- LangChain and LangGraph
- Fine-tuning
- Open-source models with QLoRA
- Deployment
- FastAPI services
Selected work
A fine-tuned 1.3B model outperformed a GPT-4o baseline.
Master's dissertation, Cardiff University, Jun–Sep 2025 (High Distinction, 81%). Small open-source models were fine-tuned to answer questions over tables.
LLM Data Agents for Tabular Reasoning: pipeline
- DataBenchThe benchmark the dataset is built from.
- Curate and verifyAutomated pipeline for (Question, Thought Process, Code) triplets, verified programmatically.
- QLoRA fine-tuningMethod tailored to each architecture.
- DeepSeek-1.3B
- Phi-3-3.8B
- CodeQwen-7B
- EvaluateUnseen test set, compared with a GPT-4o baseline.
Autonomous Financial Deep Research Agent
Financial AI Infrastructure, Jun–Sep 2026
LLMs make arithmetic mistakes, and fixed parallel searches cannot adapt to what earlier steps found.
- Tool-calling agent in Python with an iterative search loop, programmatic step-nudge and source-discrepancy budget controls.
- RAG pipeline using ChromaDB and Google's gemini-embedding-001 for semantic search over local company annual reports.
- Separate Python calculation layer for YoY growth and operating margins, with safeguards against conflating EBIT and EBITDA margins.
- Python
- RAG
- ChromaDB
- LangGraph
Malware Analysis Using Deep LearningMahindra University, Aug–Dec 2023
A malware detection model in Python, TensorFlow and Keras using BERT and Word2Vec embeddings, iteratively evaluated and fine-tuned to improve classification accuracy.
NLP Emotion AnalysisMahindra University, Jan–May 2023
A HuggingFace-based service that analyzes and compares the emotional depth of AI-generated stories, with emotion probability matrices and bar charts.
Experience
- Jan 2024 – May 2024
Intern, Digital Innovator
Didgiup Private Limited- Designed and built a multi-output classification model for a Door Specification project using decision trees and neural networks.
- Deployed a production-ready FastAPI web service to process input data, predict classifications and handle document uploads.
- Used pdfplumber and python-docx for data extraction pipelines, and a TfidfVectorizer to train Random Forest and AdaBoost classifiers.
- Jun 2023 – Aug 2023
Data Acquisition for Smart Ankle-Foot Device
Research project, IIT Jammu- Engineered a Python data acquisition system to collect and process EMG sensor data.
- Implemented and deployed machine learning models to classify and predict motor actions from real-time EMG signals, a full cycle from data to deployment.
Two degrees in AI.
Cardiff, UK
MSc Artificial Intelligence
Cardiff University. Distinction (overall 75%). Dissertation: High Distinction (81%).
Hyderabad, India
B.Tech Artificial Intelligence
Mahindra University.
Skills
- Language
- Python (proficient)
- AI / ML
- LangChain
- LangGraph
- ChromaDB
- TensorFlow
- Keras
- Scikit-learn
- HuggingFace
- Pandas
- NumPy
- Deployment
- FastAPI
- Streamlit
- Git
- Competencies
- Agentic AI
- RAG
- LLM Orchestration
- NLP
- Deep Learning
- API Deployment