Headline

GenAI Engineer | NLP & LLM Applications | RAG | Agentic AI | Python

About

I’m a GenAI Engineer with 3.5 years of experience across Machine Learning, Natural Language Processing, and Generative AI, with experience spanning fintech, healthcare/clinical NLP, and data-driven AI applications. I specialize in building LLM applications, RAG and Agentic RAG systems, NLP pipelines, semantic search, and transformer-based models, using Python and modern AI frameworks My work sits at the intersection of NLP and AI engineering, from information retrieval and document intelligence to LLM-powered applications and domain-adapted transformer models. I’ve worked with LangChain, LangGraph, GPT-4o, FAISS, Sentence Transformers, BERT-family models, LoRA fine-tuning, FastAPI, and vector search, with a focus on retrieval quality, evaluation, reliability, and practical AI applications. I’m particularly interested in opportunities where I can work on GenAI, NLP, LLMs, RAG, Agentic AI, information retrieval, and AI application engineering especially problems where model quality, retrieval, evaluation, and system reliability matter as much as the underlying model. Core areas: Generative AI · NLP · LLMs · RAG · Agentic AI · Transformers · Fine-Tuning · LoRA · Semantic Search · Information Retrieval · Vector Search · LangGraph · LangChain · Python · FastAPI

Location

Bengaluru

AI stack

claude code

cursor

github copilot

hugging face

langgraph

opencode

Work experience

Data Science Intern at Infyntrek Systèmes, Apr 2026 - Jul 2026 · 3 mo, Remote, Built a production-grade text-to-SQL system that blocks 100% of destructive queries via a sandboxed guardrail, detects 50% of hallucinated answers before they reach users through back-translation, sanity, and confidence checks, and hits 96.3% execution accuracy on a 54-case benchmark.

Technical Consultant at Freelance, Feb 2026 - Apr 2026 · 3 mos, Maharashtra, India, Built 7 public pages: Home, About, Courses, Exams, Results, Gallery, Contact fully responsive and mobile-first integrated 46 Radix UI components via shadcn/ui for accessible consistent UI patterns (dialogs, forms, tables, carousels, toasts). Used Framer Motion for page transitions, parallax effects, and animated stats counters. Built a RESTful API with 8 endpoints served by Uvicorn ASGI Used Pydantic v2 for strict request/response validation catching malformed data at the API boundary before it hits the database Connected to Supabase PostgreSQL via async supabase-py client with JSONB columns for flexible data. Added slowapi rate limiting and fastapi-mail email notifications triggered on every inquiry submission Wrote API integration tests using Pytest + httpx to validate all endpoints

Research Assistant at University of Nottingham, May 2025 - Sep 2025 · 5 mos, Nottinghamshire, Constructed OpenI-NLI, a novel silver-labeled radiology NLI corpus from 3,412 real clinical reports, applying BM25 retrieval and negation-polarity heuristics for annotation-free 3-class labeling, establishing a new research direction for data-driven NLP in medical text analytics. Domain-adapted BiomedBERT via LoRA (rank=16, 0.45% trainable params) on a single T4 GPU, achieving 0.978 F1 on entailment across 672 held-out radiology pairs, and deployed a hallucination detection pipeline (FastAPI + SQLAlchemy, React frontend) producing per-claim verdicts with confidence scores and evidence pointers

Python Developer at Hexaware Technologies, Feb 2022 - Jul 2024 · 2 yrs 6 mos, Pune District, Built an end-to-end document classification system (TF-IDF + LinearSVC, Scikit-learn) for a fintech client, achieving 91% accuracy across 7 document types, delivered a working prototype in 5 days from ambiguous requirements, securing a 6-month $200K client engagement. Scaled classification accuracy to 94% by integrating Tesseract OCR and fine-tuning a LayoutLM model for spatial-aware field extraction across invoices, POs, and KYC forms, reducing manual document sorting workload by 70%, selecting LinearSVC over deep learning alternatives via stratified 5-fold cross-validation for explainability and deployability

Education

Master of Science, Computer Science, University of Nottingham