Sahil Bansal

AI Researcher

San Francisco, California, United States6 yrs 3 mos experience

Most Likely To SwitchAI Enabled

Key Highlights

Expert in Generative AI and Knowledge Graphs.
Proven track record in building advanced NLP models.
Strong background in machine learning and deep learning.

Stackforce AI infers this person is a specialist in AI and machine learning for enterprise applications.

Contact

sahil.bansal01@sap.com +919878253415 LinkedIn

Skills

Core Skills

Generative AiKnowledge GraphsRetrieval-augmented Generation (rag)Natural Language Processing (nlp)Machine Learning

Other Skills

AlgorithmsBashC++DatabasesDeep LearningDigital Signal ProcessingGitInformation RetrievalLaTeXLarge Language Models (LLM)ProbabilityPyTorchPythonResearchSQL

Experience

6 yrs 3 mos

Total Experience

1 yr 6 mos

Average Tenure

2 yrs 1 mo

Current Experience

Sap

Senior Data and Applied Scientist, Generative AI and Knowledge Graphs

Apr 2024 – Present · 2 yrs 1 mo · Palo Alto, California, United States · Hybrid

Spearheaded intelligent semantic segmentation for SAP J4C retrieval, enhancing relevancy, faithfulness, and certification exam precision
Developed a knowledge graph–based tool retrieval framework for AI agents with 91.85% tool coverage (vs. 89.26% for similarity-based approaches)
Leading end-to-end training & fine-tuning of SAP’s Embedding Model, covering data generation, benchmarking, training, and evaluation for product integration
Building a training framework for SAP-specific language models, leveraging synthetic QA generation to support answer generation and knowledge modeling
Planning Agents on an Ego-Trip: Leveraging Hybrid Ego-Graph Ensembles for Improved Tool Retrieval in Enterprise Task Planning: https://arxiv.org/pdf/2508.05888 (IJCNLP-AACL 2025 Findings)
Towards Practical GraphRAG: Efficient Knowledge Graph Construction and Hybrid Retrieval at Scale: https://arxiv.org/pdf/2507.03226 (Under Review)

Large Language Models (LLM)Knowledge GraphsInformation RetrievalRetrieval-Augmented Generation (RAG)Word EmbeddingsSynthetic Data Generation+1

Hive

Machine Learning Engineer

May 2023 – Mar 2024 · 10 mos · San Francisco, California, United States · On-site

Built a production-ready text moderation model for enterprise customers using large-scale pre-training, and distillation of BERT-like models
Developed and improved instruction-tuned large language models for advanced assistant applications
Worked on the low-rank adaptation of large language models (LoRA), providing support for extended context scenarios in assistant applications

Deep LearningText ModerationNatural Language Processing (NLP)Large Language Models (LLM)Machine Learning

University of california, los angeles

3 roles

Graduate Teaching Assistant

Sep 2022 – Mar 2023 · 6 mos · Los Angeles, California, United States · On-site

CS 181 - Introduction to Formal Languages and Automata Theory

Graduate Student Researcher

Apr 2022 – May 2023 · 1 yr 1 mo · Los Angeles, California, United States · On-site

Peng’s Language Understanding & Synthesis (PLUS) Lab

Graduate Reader

Jan 2022 – Jun 2022 · 5 mos · Los Angeles, California, United States · On-site

Data Science - CS M148, Artificial Intelligence - CS 161, Algorithmic Machine Learning - CS 260B

Amazon

Applied Scientist Intern

Jun 2022 – Sep 2022 · 3 mos · Seattle, Washington, United States

With Payments Acceptance and Experience - Data Science Team (Applied Science and Machine Learning Engineering) under the World Wide Consumer Payments division at Amazon.

Ibm

Research Software Engineer

Aug 2019 – Aug 2021 · 2 yrs · New Delhi Area, India

Built an out-of-the-box system for detecting log anomalies and sending real-time alerts using error classification, log templatization, and template clustering models. Work got incorporated in IBM Cloud Pak for Watson AIOps.
Developed an end-to-end system to perform automatic document classification and clustering for machine learning-based document processing systems.
Created a pipeline for extending a knowledge graph for IT Operations using the glossaries available in the IT domain. Pipeline made use of propositionalization techniques, text-based embeddings, clustering, and graph-based approaches.
Worked on automatic extraction of component-action relationships and other related attributes from support queries using a shallow semantic parsing-based approach.

Mila - quebec artificial intelligence institute

Visiting Researcher

May 2019 – Jul 2019 · 2 mos · Montreal, Canada Area

Worked with the Climate Change AI group at MILA, supervised by the Turing Laureate Prof. Yoshua Bengio and Dr. Sasha Luccioni
Goal was to use generative approaches to produce images of how neighborhoods and houses will look like following the effects of a calamity
Solution built upon the use of generative adversarial approaches, water segmentation models and domain adaptation techniques for generating realistic future predictions for extreme weather events

Adobe

Research Intern

May 2018 – Jul 2018 · 2 mos · Bangalore

The project aimed at looking for different ways of assisting users of search systems in constructing richer queries more likely to retrieve desired results
The research problem that was the focus of the internship was to use historical data to build machine-learned models that generate reformulation suggestions
Solution built upon the use of generation-based Seq2Seq model for capturing session context, and a multi-task architecture for optimizing the ranking of results

Indian institute of science (iisc)

Research Intern

May 2017 – Jul 2017 · 2 mos · Bangalore, India

Worked on Adaptive Frequency Estimation using Iterative DESA with RDFT-based Filter
Combined Discrete Energy Separation Algorithm (DESA-1) and RDFT-Based Filter to design a new technique for Grid Voltage Frequency Estimation
Proposed technique is computationally efficient and robust to negative effects caused by noise and harmonics