Headline

Senior Data Scientist | MSc Data Science and Analytics, Leeds | LLM’s | Statistical Analysis | Data Management | Ex-Informatica

About

As a Data Science and Gen AI Engineer, I design and deploy LLM powered agentic systems and scalable ML pipelines that turn cutting edge AI research into production ready applications. With over four years of experience across DataINFA, NHS Digital (NHSx), and Informatica, my work bridges the gap between intelligent automation, data engineering, and applied machine learning. At DataINFA, I lead the development of multi agent orchestration frameworks using LangChain and CrewAI, enabling self governing data stewards that achieve near real time compliance and decision automation across large scale datasets. My recent projects include fine tuning Llama 3 and 4 models, building RAG pipelines for document intelligence, and deploying LLM agents that enforce data governance policies with measurable precision. Previously, I built production ML pipelines for NHS Digital, automating synthetic data generation and evaluation for healthcare analytics at scale. My foundation in computer science and MSc in Data Science and Analytics from the University of Leeds sharpened my analytical depth and ability to translate technical rigor into business impact. I am passionate about agentic AI systems, retrieval augmented reasoning, and evaluating LLM reliability in real world environments, pushing AI from experimentation to scalable, trusted infrastructure.

Location

Bengaluru

AI stack

claude

cursor

langchain

langgraph

openai

Firecrawl

Work experience

Lead Data Scientist at DatAInfa | DFactory I DINFA, Oct 2024 - Present · 1 y 11 mo, Bengaluru, ● Architected an Hybrid AI Data Steward system leveraging LangChain and CrewAI for multi-agent orchestration to automate data change request (DCR) governance. Designed specialized agents for validation, compliance checks, and automated decision-making across 10+ business rules achieving 95% audit success and 85% reduction in manual review time, while processing thousands of records per hour using fuzzy matching (Damerau-Levenshtein distance) and international address validation across 50+ countries. ● Engineered a production-grade OCR + LLM + RAG pipeline using FastAPI, LangChain, and FAISS to convert 22k+ unstructured documents into structured MDM entities. Deployed on scalable cloud infrastructure supporting 50+ concurrent requests with <500ms P95 latency. ● Fine-tuned Llama 4 (maverick) using QoRA and RAG architectures for domain-specific retrieval from Informatica Knowledge Base articles. Implemented a systematic evaluation framework that improved search accuracy by 50%+ and achieved a 95% relevance score through iterative optimization. ● Developed LLM-based compliance automation agents with advanced prompt engineering for regulatory adherence (HIPAA, GDPR, DPDP). Deployed production systems handling 1K+ policy assessments monthly with real-time performance monitoring. ● Built robust evaluation pipelines featuring custom metrics and benchmark suites to validate agent behavior and ensure production reliability, reducing false positives by 20%. ● Drove rapid prototyping cycles, consistently transitioning features from proof-of-concept to production within 2–4 weeks, supported by thorough documentation of design trade-offs and technical decisions with stakeholders. Key Technologies: CrewAI, LangChain, PyTorch, Hugging Face Transformers, FAISS, FastAPI, Groq API, Azure OpenAI, RAG, LoRA/PEFT

Research Data Scientist at University of Leeds, Jul 2024 - Feb 2025 · 8 mos, Leeds, England, United Kingdom, ● Led the development of a synthetic blood transfusion dataset for NHS Blood Transfusion unit funded by NIHR, collaborating with University of Oxford clinicians, doctors, and project managers to generate 50,000+ anonymized records using statistical and Deep Learning models. ● Embedded clinical insights into data pipelines to simulate real-world transfusion scenarios (adverse reactions, blood product usage) while ensuring strict privacy safeguards and compliance with NHS governance standards. ● Built an open-source tool adopted by cross-functional teams to generate and validate synthetic data, enabling ethical research into rare transfusion outcomes and policy testing without compromising patient confidentiality.

Dissertation Partner at NHS, Mar 2024 - Sep 2024 · 7 mos, Leeds, England, United Kingdom, ● Generated synthetic Electronic Health Records using Normalising Flows and GANs by developing and training these models to simulate patient data (MIMIC-IV), reducing reliance on real patient data by 75%. ● Achieved a 95% statistical match using the Kolmogorov-Smirnov test between synthetic and real data, ensuring high fidelity while complying with GDPR regulations. ● Deployed the model on AWS SageMaker to automate the generation of synthetic medical data for the open-source medical record system (OpenMRS) platform. ● Presented insights to NHS stakeholders (Senior Data Scientists, Data Stewards, GPs) through monthly reports, leading to iterative improvements and alignment with key NHS project goals.

Informatica Data Engineer - Data Consultant at Informatica, Apr 2023 - Sep 2023 · 6 mos, Bengaluru, Karnataka, India, ● Implemented Master Data Management (MDM) solutions by designing and deploying the Informatica cloud product suite; fine-tuned the SSA-NAME3 algorithm to detect 3 million duplicates improving data consistency and integrity across the enterprise. ● Automated data profiling using Pandas Profiling to assess data quality of large datasets, enhancing efficiency in data quality analysis by integrating this solution into Informatica Cloud for seamless deployment and scalability. ● Collaborated with Data Engineering, DevOps, and Customer Success teams to communicate insights and model performance, ensuring seamless adoption and alignment with business objectives. ● Provided expert consultation acting as SME to Fortune 100 clients on data analytics and quality, introducing best practices for advanced analytics techniques and data science.

Associate Consultant at Informatica, Aug 2020 - Apr 2023 · 2 yrs 9 mos, Bengaluru, Karnataka, India, Active Development of the MDM Healthcheck Tool which is used to assess the health condition of the MDM System. The health check includes a report on the MDM configuration and recommendations on the implementation part. Technologies used: JAVA (Application), GROOVY Engine (Rule Generation), PYTHON (Recommendation), DTE Engine (Report Generation).

Intern - Master Data Management at Informatica, Jan 2020 - Jul 2020 · 7 mos, Greater Bengaluru Area, Worked on the Master Data Management project using Informatica MDM (Multi-Domain Edition), solution, which binds systems & information together i.e. business data that spread across over industry.

Summer School (on campus) at Indian Institute of Technology, Delhi, Jun 2018 - Jul 2018 · 2 mos, Greater Delhi Area

Education

Msc, Data Science and Analytics (on-campus), University of Leeds

Bachelor of Engineering - BE, Computer Science, KLE Technological University - Hubballi (India)

Profile highlights

What I'm currently working on: At DataINFA, I am building multi-agent systems using LangChain to automate data governance. I am currently fine-tuning Llama models to create self-governing stewards that enforce compliance in real-time.

The last shipped feature I'm proudest of…: I recently deployed a multi-agent orchestration framework that automates data governance at scale. By integrating LangChain agents, I enabled self-governing data stewardship that achieves near real-time compliance across our infrastructure.

My hot take about working with AI is…: LLMs are useless without strict, automated data governance to prevent them from hallucinating errors.

Public links

github: https://github.com/vellalavineethkumar