About

Hello, my name is Sankalok (pronounced shaun-ko-lock). I like thinking deeply about complicated problems, writing down and discussing possible solutions with colleagues, and tinkering with computers/servers while fully using up all of available resources. I also try to publish once in a while.

Experience

Huawei Technologies

Hong Kong

L15 • Engineer AInterdisciplinary Technology Innovation CenterDec 2026 – Present

L14 • Engineer BTheory LabFeb 2024 – Dec 2026

  • Led a team of three interns on prefill and decode optimization for text and multimodal embedding, reranking, and OCR models, and on automated NPU kernel-generation pipelines for Data Storage and Consumer Business Group product lines.
  • Built a serving engine on Ascend NPUs for prefill-only models that outperforms vLLM-Ascend and TEI; the approach was adopted into vLLM-Ascend for the Data Storage product line.
  • Optimized ViT-based architectures for automated feature extraction from whole-slide pathology images in Huawei ModelEngine V26.0, released as an open-source collaboration with Ruijin Hospital, Shanghai.
  • Collaborated on a text and multimodal knowledge-graph extraction pipeline for retrieval-augmented generation, with hierarchical granularity, self-repairing triplets, elastic schema induction, and intent-driven entity–event disambiguation, and on a unified multi-relation graph from raw sources for the Data Storage product line and Huawei MindSDK.
  • Contributed to a top-3 SOTA in-memory ANN method, Knowledge Graph Neighbours (kgn), by adding hierarchical clustering for vector search to raise query throughput for Terminal and Cloud Storage product lines.

B.Y. Quantitative Medicine

Hong Kong

Junior Research ScientistAug 2023 – Feb 2024

Research InternDec 2022 – May 2023

  • Built a mixture-of-experts stack for cancer drug-resistance prediction, improving performance on ovarian cancer subtypes from Hong Kong hospital data to help clinicians choose chemotherapy combinations.
  • Standardized ovarian cancer datasets with nonparametric statistical modeling so genetic encodings from different hardware could be used together in later analysis.
  • Wrote documentation and presentations that explained the research workflow to external stakeholders.

The University of Hong Kong

Hong Kong

Research AssistantJul 2022 – May 2023

  • Built an information-retrieval pipeline for Hong Kong legal judgments: converting HTML to Markdown, then keyword extraction, topic modeling, and transformer summarization to surface post-hoc insights and retrieve related rules.

Lora Technologies

Hong Kong

Research AssistantJun 2022 – Jul 2023

  • Designed a news-scouting ETL pipeline that scored article relevance, dropped stale items, and migrated results into ScyllaDB for faster column-oriented analysis.
  • Built a regression-based market-sentiment indicator for equities and indexes from in-house news data, to study retail trading activity after the fact.

Saint Mary's University

Halifax, Canada

MITACS Globalink Research InternJul 2021 – Oct 2021

  • Built a pipeline to clean Fortune 500 CSR reports, extract keyword usage with topic modeling, and store the results for later analysis.
  • Visual analysis showed a clear rise in DE&I-related language in the late 2010s compared with earlier decades.

Education

The University of Hong Kong

Hong Kong

BEng in Computer Science, minor in StatisticsSep 2019 – Jun 2023

  • Teaching Assistant: COMP1117 / ENGG1330, Introduction to Programming in Python, Fall 2020 and Spring 2021.

Awards

Publications

  1. Yazheng Yang, Yuqi Wang, Yaxuan Li, Sankalok Sen, Lei Li, Lin Qiu and Qi Liu, Unlock the Potential of Large Language Models for Predictive Tabular Tasks in Data Science with Table-Specific Pretraining. IEEE TKDE 38(6), June 2026.
  2. Sankalok Sen, Analyzing Hong Kong’s Legal Judgments from a Computational Linguistics point-of-view. arXiv, May 2023. Undergraduate thesis.
  3. Srinjoy Bhuiya, Ayushman Kumar, and Sankalok Sen, Exploring the Effects of Data Augmentation for Drivable Area Segmentation. arXiv, August 2022.

Disclaimer: Opinions are my own and do not reflect those of my employer.