KSHITIJ KAVIMANDAN
Data Engineer
Publication
"Patent Retrieval with Hierarchical Multi-Positive Contrastive Learning" presented at SIGIR 2025 in Padua, Italy—one of the top conferences in information retrieval.
Education
M.Sc, Computer Science
University of Amsterdam 2023 - 2025SKILLS
- Azure
- Docker
- PyTorch
- CV
- ETL
- Kubernetes
- LLMs
- NLP
Experience
AI Research Engineer
Sept 2024 – March 2025 | TKH Artificial Intelligence, AmsterdamImproved patent image retrieval accuracy by 110% by implementing hierarchical multi-positive contrastive learning leveraging Locarno classification embeddings.
Software Engineer
May 2022 – June 2023 | Gupshup, MumbaiBuilt and deployed over 25 client-specific chatbots using Node.js and MongoDB, integrating APIs for seamless customer interactions across various social media platforms.
Projects
Efficient Data Pipeline with Incremental Updates (2024)
Developed Python-based incremental view maintenance system (pg-ivm), enhancing PostgreSQL query speed by 200%, significantly cutting computation time over full recomputation using HyperLogLog algorithms.
Containerized Microservices Architecture (2024)
Built scalable microservices on Kubernetes with automated CI/CD pipelines, custom Helm charts, TLS encryption, and RBAC policies, reducing deployment time by 40%
Factyr - LLM Response Enhancer (2023)
Implemented entity linking and knowledge-base fact verification modules with spaCy and Wikidata, achieving 30% reduction in LLM hallucinations and enhancing factual accuracy.
Languages
- Python
- C/C++
- JavaScript
- SQL
- R