Experience

The career so far

Four years of NLP, retrieval and making analysts' lives easier.

Jun 2026 - Present

Data Scientist

J.P. Morgan · London

New chapter, bigger models. Stories loading…

  • Just joined. The interesting war stories are still somewhere between onboarding and NDA. Check back soon.

Sep 2024 - Jun 2026

Associate Data Scientist

FinTru · London

Built AI agents that read compliance documents so humans don’t have to.

  • Devised a KYC profiling AI agent with a human-in-the-loop workflow that generates 300+ risk profiles a month and cuts analyst effort by 75%.
  • Fine-tuned LLMs (Falcon, Llama 2, Qwen) with LoRA/QLoRA for retriever-reader QA systems, benchmarked against in-house models.
  • Architected memory-efficient RLHF/DPO training pipelines with 4-bit quantization on single-node A100s.
  • Authored a white paper on synthetic data generation for a proposed FCA partnership.

Sep 2022 - Sep 2024

Data Scientist

FinTru · Belfast / London

Made retrieval fast and deployment decisions obvious.

  • A/B tested BM25, DPR and ColBERT retrievers, cutting inference latency by 50%+ and saving ~2000 analyst hours a year.
  • Built the model performance dashboard (Power BI) used across 8 production models, saving 100+ hours a month.
  • Integrated PyTorch dataloaders + MLflow experiment tracking; hyperparameter search lifted F1 by 15%.
  • Ran the data science internship programme and mentored a master’s thesis.

Jun 2022 - Aug 2022

NLP Data Science Intern

FinTru · Belfast

The summer that started the whole NLP thing.

  • Built 4 text-categorization models (Naive Bayes → SVM), with a rule-based pre-processing framework for imbalanced data.
  • Reached 97%+ categorical accuracy with BERT; validated with 5x2 cross-validation statistical tests.

Jan 2020 - Mar 2020

Data Science Intern

Ernst & Young India · New Delhi

First taste of enterprise data: SAP, SQL and a lot of dashboards.

  • Extracted data from SAP/SQL databases with Denodo for the Risk Advisory team.
  • Cut query run-time 30% with scheduled data caching; shipped 3 real-time Power BI dashboards.

Education

2021 - 2022

London School of Economics

M.Sc. Operations Research & Analytics

Algorithms & Computation, Data Mining, Deep Learning, Modelling & Simulations

2017 - 2021

NIT Kurukshetra

B.Tech Production & Industrial Engineering

Operations Research, Supply Chain Management, Machine Design

The toolbox

LLMs & NLP

LLM Fine-tuningLoRA / QLoRARLHF & DPONERRetrieval (ColBERT, DPR, BM25)Hugging FaceMulti-Agent Systems

Data & ML

PythonPyTorchSQLPandasScikit-learnA/B TestingPower BI

Infra & Tooling

AzureDatabricksMLflowDockerKubernetesStreamlitGit