Jay Singh

AI Researcher

India3 yrs 10 mos experience
Most Likely To SwitchAI ML Practitioner

Key Highlights

  • Ranked among the top 7% of Kaggle Notebooks Expert community.
  • Engineered an intelligent Intrusion Attack Detection System with 85.45% accuracy.
  • Developed an AI-driven Startup Portal enhancing decision-making efficiency by 40%.
Stackforce AI infers this person is a Data Science and Machine Learning specialist with a focus on advanced NLP and AI-driven solutions.

Contact

LinkedIn

Skills

Core Skills

Federated LearningIntrusion Detection SystemPredictive AnalyticsData VisualizationTeam LeadershipProject ManagementMarketing StrategyNatural Language Processing (nlp)TransformersLarge Language Models (llm)Generative Ai

Other Skills

AI AgentsAgentic WorkflowAmazon Web Services (AWS)AnacondaAnthropic ClaudeApplied MathematicsArtificial Intelligence (AI)Attention MechanismsAutomatic Text SummarizationBERT (Language Model)Bayesian OptimizationBusiness StrategyComputer EthicsContext ManagerConvolutional Neural Networks (CNN)

About

Hello there 👋🏻, I am a final-year undergraduate at IIT Kharagpur, driven by a passion for leveraging advanced machine learning to solve real-world challenges. Starting my Python journey 4year back, I continued to learn advanced tech stack (both using Python and beyond), built real world projects, gradually stopping into ML/DL. My experiences at Innovaccer, Affine and other companies have given me hands-on experience in transformer-based models and developing multi-agent systems for dynamic language processing. I excel in Python, TensorFlow, PyTorch, and Hugging Face, and have utilized tools like MLflow, LangChain, Streamlit to streamline complex ML workflows. Beyond my formal experience, I am an active Kaggle contributor, consistently tackling challenging data problems and deliver innovative solutions. From covering intermediary domains like NLP, Sentiment analysis, Time-Series, etc to advanced use cases utilising Transfer Learning, Federated Learning (ZSL), Agentic Workflows, GNNs, etc. I commit to excellence and continuous learning in the competitive landscape of data science. In addition, I have built and deployed robust web applications incorporating Django, Flask, and FastAPI, and managed scalable cloud solutions on AWS with Snowflake for efficient data handling. My leadership as a cultural fest coordinator has further refined my strategic planning, teamwork, and communication skills. Combining technical mastery with a proven track record in competitive environments, I am eager to bring my diverse expertise and proactive mindset to dynamic teams, driving impactful innovations and delivering high-quality solutions in data science and machine learning.

Experience

3 yrs 10 mos
Total Experience
1 yr 4 mos
Average Tenure
2 yrs 7 mos
Current Experience

Indian institute of technology, kharagpur

Research Project

Dec 2024 – Apr 2025 · 4 mos · On-site

  • Title - Enhancing Intrusion Detection System using Federated Learning
  • Engineered an intelligent Intrusion Attack Detection System utilizing Zero Shot Federated Learning approach, with an unknown attack prediction accuracy of 85.45%.
  • Engineered refined prompts for Llama-7B-chat model to generate unique quality attack description for each corresponding data entry, incorporating into the training dataset of 60k size.
  • Built a rich corpus of IDS attacks, along with descriptions and 768 dimentional vector embedding of each, with the help of CAPEC Mitre resource, building a massive library of attacks, helping in model response testing.
  • Created a multi hidden Densely connected Neural Network with over 70k parameters under Federated learning architecture of client - server internal weight exchange pipeline.
  • A robust architecture optimized by Bayesian Optimization method, delivered unmatched results detecting unknown test features.
Federated LearningIntrusion Detection SystemDeep LearningFeature EngineeringBayesian Optimization

Symx.ai

Data Scientist

Sep 2024 – Feb 2025 · 5 mos · Hybrid

Predictive AnalyticsAmazon Web Services (AWS)Multivariate StatisticsData Visualization

Spring fest

3 roles

Steering Committee Member

Jun 2024 – Jan 2025 · 7 mos

Publicity and Media Outreach Head

Jun 2023 – Jun 2024 · 1 yr

Team LeadershipPublic RelationsStrategic ThinkingProject Management

Core Organising Team member

Aug 2022 – Jun 2023 · 10 mos

  • Spearheaded Sponsorship drive in Uttar Pradesh raising INR 1 lakh+ (200% YoY increase) through alumni contribution
  • On-Ground experience in conducting events in the cities of Lucknow, Patna as the Event organiser.
  • Spearheaded publicity drive in UK, Goa, Punjab, witnessing 50% YOY in participation.
  • Spearheaded a team of 4 members to successfully secure 26 media partners for Spring Fest in the
  • Publicity and Media Outreach domain, including deals with esteemed outlets such as Lutopia
  • Magazine, Punjab Kesari, Amar Ujala and more.
  • Successfully conducted SF Talkies with guest being the proficient bollywood actor Amit Sadh.
  • Supervised and executed the 4-day pronites, featuring renowned artists like Nikhil D'Souza, King,
  • Nucleya and Sunidhi Chauhan, managing a massive footfall of over 50,000.
Marketing StrategyGrowth StrategiesProject ManagementCritical Thinking

Innovaccer

Data Science Intern

Jun 2024 – Aug 2024 · 2 mos · Noida, Uttar Pradesh, India

  • Worked on the natural language semantics of "Sara’s", AI model of Innovaccer, with a progression from NLTK to transformer based architecture, as part of Innovaccer’s NLP R&D team.
  • Explored open-source models like GTR-T5 base, Flan-T5 XL, and others. Fine-tuned models over single, multi-GPU processing for optimization, for refined outputs and low computation time, respectively.
  • Leveraged LangChain and diverse prompt engineering techniques, including Chain of Thought, Zero-shot, and Few-shot learning leading to high precision results.
  • Built a RAG pipeline over gpt 4o, Milvus as the vector database, and gte-large, COLBERT for generating multi-vector embedding for high precision, fast retrieval.
  • Built a unified asynchronous model chaining and dividing the multiple task focused model on top of lexical similarity architecture.
LangChainTransformersNLTKRetrieval-Augmented Generation (RAG)sbertPyTorch+3

Affine

Machine Learning Intern

Apr 2024 – Jul 2024 · 3 mos

  • Partnered with cross-functional engineering teams to architect and deploy a custom T5-based pipeline, selecting a 220 M-param model (12 attention heads, hidden size 768) for optimal cost–performance trade-off. Leveraged mixed-precision (FP16) and gradient checkpointing to slash GPU memory usage by ~55 %—enabling training on 4 Ă— A100 40 GB nodes instead of 8.
  • Adopted Low-Rank Adaptation (LoRA) for instruction fine-tuning, injecting just 2 % extra parameters (~4 M) to specialize the model on domain-specific prompts. This approach delivered a 1.7 BLEU / ROUGE-L point lift while reducing fine-tuning GPU hours by 40 %, compared to full-parameter updates.
  • Integrated 8-bit quantization (bitsandbytes + ONNX Runtime) in inference, compressing the model by 4Ă— on disk and boosting throughput by 3×—achieving sub-200 ms latency on GPU and <500 ms on CPU for 100-token generative requests, all with <1 % drop in generation quality.
  • Engineered an end-to-end MLOps pipeline using Kubeflow, MLflow, and Triton Inference Server, enabling dynamic batching, auto-scaling, and real-time monitoring. This infrastructure supported 10Ă— user growth without additional engineering overhead and drove a 30 % reduction in cloud costs through spot-instance utilization and optimized resource scheduling.
Large Language Models (LLM)Natural Language Processing (NLP)TransformersGenerative AIFine Tuning

Kaggle

Kaggle Expert

Nov 2023 – Present · 2 yrs 7 mos

  • Kaggle Rank: Ranked among the top 7% of Kaggle Notebooks Expert community (3,947 of 61,474).
  • Community Contribution: Published a ton of notebooks over diverse Problem Statements corresponding to different domains of ML/DL like NLP, Classifications Techniques, Regression Models, fine-tuning of LLMs, many more.

Cambridge judge business school

Research Associate

May 2023 – Sep 2023 · 4 mos · United Kingdom

  • Developed an AI-driven Startup Portal using Groq API, delivering personalized strategies, business ideas, and enhanced user engagement
  • Engineered NLP techniques for extracting key insights from inputs, generating SWOT analyses, boosting decision-making efficiency by 40%.
  • Integrated prompt engineering, enabling context-aware output, addressing marketing strategies, financial planning, resource optimization
  • Built scalable RESTful endpoints with robust data parsing and CORS enabling seamless and real-time interactions across multiple platforms
GroqGenerative AIRESTful WebServicesNatural Language Processing (NLP)React Native

Education

Indian Institute of Technology, Kharagpur

Jan 2021 – Jul 2026

Jay Singh - AI Researcher | Stackforce