AI · Data · Software Engineering

Hi, I'm Nishit Thakkar.

I build end‑to‑end intelligent systems – combining machine learning, data engineering, and backend development. My experience spans knowledge graphs, RAG systems, multimodal computer vision, and production SaaS backends. I am based in London, UK and open to roles that blend AI/ML, data, and software engineering.

View Projects View Experience
  • Right to Work: UK Graduate Route / Post‑Study Visa
  • Location: London, United Kingdom
Email · LinkedIn · GitHub · Download CV
Portrait of Nishit Thakkar

Profile at a glance

  • Blend of AI/ML, data engineering, and backend development.
  • Built production RAG systems processing 4M+ PubMed papers.
  • Engineered knowledge graphs with 88K+ nodes & 824K+ edges.
  • Delivered custom data‑driven solutions for 8+ enterprise clients.

Summary

I am an engineer who enjoys working across the full stack of intelligent systems – from data ingestion and modelling to ML research and backend deployment. My work includes graph‑based drug repurposing pipelines, retrieval‑augmented generation over millions of biomedical documents, and production‑ready SaaS features for enterprise clients. I bring together AI/ML, data science, and software engineering to deliver systems that are robust, interpretable, and useful in the real world.

Skills

Core Programming

Python (Advanced), SQL, Bash, JavaScript, HTML, CSS, Git

Machine Learning & AI

Scikit‑learn, PyTorch, TensorFlow, Applied ML, Model Evaluation, Gaussian Splatting, Computer Vision, NLP, Knowledge Graphs, RAG Systems, Explainability

Data & Backend Engineering

Data Pipelines, ETL, Data Scraping, Data Modelling, REST APIs, Django, Flask, Odoo ERP, PostgreSQL, MySQL

Tools & Delivery

Docker, Linux, CI/CD, Agile Development, Experimentation, SaaS Environments, Client‑facing Delivery

Experience

Data Scientist / ML Engineer

Topia Life Sciences, India

Mar 2024 – Jan 2025

AI · Data · Backend

  • Designed and maintained large‑scale data pipelines to ingest and process biomedical data from clinical trials, PubMed, and research repositories.
  • Engineered a production‑grade knowledge graph with 88K+ nodes and 824K+ edges to model drug relationships, improving link‑prediction accuracy from 65% to 88%.
  • Implemented graph algorithms and predictive models that identified 9 viable drug‑repurposing candidates over a 10‑month period.
  • Built an automated Retrieval‑Augmented Generation (RAG) system processing 4M+ PubMed papers and 100K+ clinical trials, reducing expert research time from weeks to hours and to ~20 hours per query.
  • Enhanced explainability of model outputs using classical graph traversal techniques, enabling transparent, research‑grade decision support.

Software Developer (Data‑Focused)

Odoo, India

Jan 2023 – Mar 2024

SaaS · Backend · Data

  • Delivered end‑to‑end custom Odoo solutions for 8+ enterprise clients, designing and deploying tailored modules, workflows, and data‑driven features.
  • Developed backend systems and APIs in Python and PostgreSQL to automate data processing and business operations across multiple SaaS deployments.
  • Optimised data‑heavy processes and resolved performance bottlenecks in large production systems, improving reliability and latency.
  • Automated a manual view‑fixing process (3–4 days per request) with a heuristic‑based script, reducing developer effort and production errors.
  • Integrated an internally hosted ML‑based image‑generation service into Odoo via REST APIs, demonstrating practical end‑to‑end ML deployment.

Selected Projects

Multimodal Sensor Integration in 3D Gaussian Splatting

MSc Thesis – University of Surrey · 2025

Developed a LiDAR‑augmented 3D Gaussian Splatting framework for scene reconstruction, achieving +5 PSNR in RGB and +4 PSNR in depth over state‑of‑the‑art baselines. Proposed a loss‑balancing strategy between depth regularisation and RGB, enabling 30% faster training while preserving fine geometric detail.

Python · PyTorch · 3D Gaussian Splatting · Multimodal Fusion · Computer Vision

Trifusion Former for Underwater Image Reconstruction

Research Project – 2025

Designed a hybrid attention mechanism combining deformable, global, and high‑frequency attention for underwater image restoration, surpassing existing benchmarks. Introduced a custom loss for targeted attenuation, improving feature retention and visual quality in challenging underwater imagery.

Python · Deep Learning · Transformers · Image Restoration

Document Quality Scoring System

Smart India Hackathon 2022 – National Winner

Built a deep learning‑based validation system to classify documents as blurred or clear. Used object detection to extract key regions (photo, details, QR code) and integrated the model behind REST APIs with a web interface for real‑time document quality checks.

Python · TensorFlow · YOLO · Flask · REST APIs

Education

University of Surrey

MSc – Computer Vision, Robotics and Machine Learning

Guildford, UK

Feb 2025 – Jan 2026

Awarded Distinction. Thesis: “Multimodal Sensor Integration in 3D Gaussian Splatting for Scene Reconstruction”.

Indus University

BTech – Computer Engineering (CPI: 9.44 / 10)

Ahmedabad, India

Aug 2019 – May 2023

Winner – Smart India Hackathon 2022 (National Level). Achieved the highest SPI of 10.0 across multiple semesters.

Contact

I am open to roles and collaborations that combine AI/ML, data, and software engineering. Please reach out directly using the details below.

Email: nthakkarofficial@gmail.com

Location: London, United Kingdom

LinkedIn: nishit-thakkar-67a9961a7

GitHub: github.com/Nishit-Thakkar

CV: Download PDF