Rishav Saigal — Senior Data Scientist & AI ArchitectArchitecting Autonomous Intelligence
Senior Data Scientist | Machine Learning Engineer | Generative AI / Agentic AI Engineer | AI Architect
Hi, I'm Rishav. I specialize in engineering autonomous machine learning systems. My career centers on time-series forecasting, causal inference, and deploying production-grade enterprise Agentic AI frameworks.
Academics & Research Focus
Formal graduate education, degrees, and core empirical research vectors.
Indian Institute of Technology Patna
Jul 2026 – Jun 2029 (Expected)Ph.D. in Computer Science & Engineering
Indian Institute of Technology Patna (Pursuing).
BITS Pilani
Sep 2019 – Aug 2021M.Tech in Data Science & Engineering
Birla Institute of Technology and Science, Pilani.
RCC Institute of Information Technology (MAKAUT)
Aug 2013 – Jul 2017B.Tech in Computer Science & Engineering
Maulana Abul Kalam Azad University of Technology.
Frontier Research Themes
LLM behavioral drift, adversarial robustness, and a self-coined calcification effect in multi-agent debate systems
Autonomous ML orchestration and agentic pipeline design (AutoML, self-correcting agents)
Probabilistic time series forecasting with LLMs
Production economics and efficiency optimisation of large-scale AI agent deployments
Enterprise Experience
Track record of leading ML engineering teams and architecting production AI systems.
Cognizant
PRESENTManager (Senior Data Scientist / AI Architect)
- >Enterprise AI Architecture & Strategy (SME): Served as designated Subject Matter Expert (SME) in AI Architecture Design and GenAI; spearheaded end-to-end architectural roadmaps and technical standards for enterprise-grade Agentic AI systems across complex GCP infrastructure (Vertex AI, Cloud Run, Gemini Enterprise, Agent Engine Platform, FastMCP); designed autonomous multi-agent orchestration frameworks and token-cost observability.
- >RFP Solutioning, Pre-Sales & Resource Planning: Spearheaded architectural solution design and technical response strategy for multiple high-value RFPs as a core SME; formulated tailored architecture proposals, multi-phase resource planning models, and effort estimation matrices for enterprise executive buyers.
- >Client Deliverable Management & Engagement: Led end-to-end delivery ownership from initial client requirement discovery workshops through technical architecture, executive stakeholder POC demonstrations, and production sign-off governance across healthcare, media, and enterprise domains.
- >Team Leadership, Mentorship & Talent Acquisition: Managed and mentored cross-functional teams of ML/GenAI engineers; directed campus-to-corporate mentorship programmes for college freshers and actively contributed to tech communities fostering student growth; served as primary technical panellist conducting hiring interviews for account staffing and external lateral recruitment.
- >Production AI Governance & Reliability: Enforced robust code review standards, CI/CD automated retraining workflows, and responsible AI guardrails; implemented self-correction validation loops, intent-routing architectures, and enterprise-grade reliability benchmarks across production pipelines.
Genpact
Assistant Manager
- >Team Leadership & ML Architecture: Managed and mentored a cross-functional team of 6 ML engineers; led Agile architecture reviews, sprint delivery, and end-to-end MLOps solution deployments on AWS SageMaker and cloud environments; defined engineering best practices and delivery roadmaps.
- >Executive Stakeholder Delivery: Served as primary technical point-of-contact for executive client demos and quarterly business reviews; translated commercial objectives into scalable machine learning specifications, delivering multi-million-dollar projected operational cost savings at 90%+ model accuracy.
- >Responsible AI Governance & Explainability: Architected production model interpretability frameworks using SHAP (XAI), customer segmentation, and causal inference uplift modeling; upskilled client data science teams on responsible AI governance and proactive drift monitoring.
Capgemini
Consultant
- >Technical Lead & Full-Lifecycle Delivery: Led cross-functional engineering teams of up to 15 members across supply chain, manufacturing, and enterprise service domains; directed end-to-end ML solution delivery from Agile discovery and architecture through production deployment on AWS and Azure cloud infrastructure.
- >Distributed Time Series & MLOps Architecture: Engineered scalable distributed time series and predictive maintenance pipelines on Azure Databricks and PySpark; automated feature engineering, hyperparameter tuning, distributed model selection, and continuous retraining workflows, cutting training latency from days to hours.
- >Deep Learning & Computer Vision Systems: Architected sequence volume forecasting architectures (LSTM, MLP) with Hungarian dispatch heuristics and frame-level CNN transfer learning pipelines for real-time asset detection; delivered interactive Plotly operational HUDs for executive decision-making.
Project Laboratory
Technical fact-sheets focusing on architecture, design, and approach details across 13 production & research projects.
Generative AI (GenAI) Video Generation
Agentic AI & GenAIJan 2026 – Present
Media Analytics (POC)
Time Series & MLOpsJun 2025 – Oct 2025
Price Elasticity (POC)
Causal ML & PricingJun 2025 – Jul 2025
Enterprise Analytics Chatbot
Agentic AI & GenAIJul 2024 – Dec 2025
Intelligent QA / BA Chatbot
Agentic AI & GenAIApr 2024 – Apr 2025
LLM-Driven Predictive Modelling
Causal ML & PricingJul 2024 – Sep 2024
LLM-Based Test Case Generator
Agentic AI & GenAINov 2023 – Mar 2024
Healthcare Consumer Analytics
Causal ML & PricingJun 2022 – Oct 2023
Demand Sensing MLOps
Time Series & MLOpsMar 2020 – Jun 2022
Predictive Maintenance
Time Series & MLOpsDec 2019 – Mar 2020
Service Now Ticket Analytics
Time Series & MLOpsOct 2019 – Dec 2019
Supply Chain Demand Forecasting (POC)
Time Series & MLOpsApr 2019 – Oct 2019
Retail Video Analytics (POC)
NLP & Computer VisionApr 2018 – Mar 2019
GitHub Open Source
Publicly available AI research & MLOps frameworks across 7 open source repositories.
model-router
Two-axis Claude Code skill that scores capability floor and cost exposure separately to pick the cheapest model.
View on GitHub →Publications & Research Vault
11 technical research publications and open datasets spanning Agentic AI, LLM behavioral drift, causal dynamic pricing, and production MLOps.
The Specialized Frontier: An Inquiry Into Gated AI Architectures and the Cooperative Safety Flywheel
Examines the emergence of gated AI architectures operating under vetted access, and proposes the cooperative safety flywheel where public interaction and red-teaming directly supply the empirical data needed to harden frontier models.
Claude Model Routing: Stop Scoring "Complexity." Score These Two Things Instead.
Argues that a single blended 'complexity' score breaks Claude model routing — short-but-hard prompts get routed too cheap, long-but-easy ones too expensive. Proposes scoring capability floor (a max) and cost exposure (a sum) as two separate axes instead, with a real downgrade test gating any model swap.
LangGraph Multi-Agent Architecture: Building a Self-Critiquing AI Debate System
Walks through building a multi-agent debate system in LangGraph, where AI agents critique and refine each other's reasoning. Covers state graph design, agent roles, and shared-memory coordination patterns.
LLM Drift Experiment: A Quantitative Framework for Measuring Behavioural Drift via Adversarial Multi-Agent Debate
An open research framework and dataset for quantifying how a model's persona and reasoning shift under sustained adversarial pressure, using a Pros/Cons multi-agent debate setup tracked across a set of behavioral signals.
Measuring Behavioral Drift in LLMs: 22 Signals, 5 Dimensions, and the Calcification Effect
Breaks down the 22-signal, 5-dimension measurement framework used to detect LLM drift, and introduces the 'calcification effect' — the tendency for a model's stance to grow more rigid the longer it's challenged.
LLM Drift Explained: Do AI Models Lose Themselves Under Adversarial Pressure?
An accessible explainer on LLM drift: what it looks like in practice, why adversarial multi-agent debate is a useful way to surface it, and what the early findings suggest about model consistency.
The $1.5 Million Difference: Why Benchmarks are Only 10% of the AI Agent Story
Argues that standard AI agent benchmarks capture only a fraction of what determines real-world production success, and outlines the operational factors — cost, reliability, observability — that account for the rest.
AutoML on Autopilot
A practical look at automating the machine learning pipeline end-to-end, from model selection to deployment, and where AutoML tooling still needs a human in the loop.
CognitoEDA
Introduces CognitoEDA, an agentic workflow that automates exploratory data analysis — schema inspection, statistical summaries, and report generation — using a multi-agent LangGraph pipeline.
Beyond Prediction: Generative AI's Probabilistic Future in Time Series
Explores how generative AI is shifting time-series forecasting from single-point predictions toward probabilistic, scenario-based forecasts, and what that means for planning under uncertainty.
Unlocking Time Series with LLMs: A New Era with TimeCAP
Looks at TimeCAP and the emerging class of LLM-based time-series models, examining how language-model architectures are being adapted for forecasting tasks traditionally owned by statistical methods.
Evolution of Capabilities
Scroll down to explore chronological skill acquisition and neural domain expertise.
Certifications & Badges Vault
14 verified professional credentials and digital badges spanning Agentic AI, Google Cloud ADK, Cloud Infrastructure, and Production MLOps.
Algorhythm Chronicles: Discover the human stories behind data science, from historical mathematics to AI, told through original music.
Algorhythm Chronicles
Subscribe on LinkedIn to receive bi-weekly narratives uncovering the mathematical breakthroughs, historical characters, and core concepts of AI and machine learning, brought to life with original thematic soundtracks.
Subscribe on LinkedIn →Algorhythm Chronicles
Initiate Collaboration
Reach out for enterprise AI architecture consultation, advisory roles, or speaking engagements.