Yash Thevalil

AI/ML Engineer | Applied Machine Learning & Analytical Systems | Software Engineer

I design and own production-grade machine learning, analytical and backend systems that operate under real-world constraints, where reliability, traceability, and long-term performance matter.

With over 5 years of experience across Europe and Asia, I work end-to-end at the intersection of AI, software engineering, and cloud infrastructure, delivering systems that hold up beyond experimentation.

Services

I help teams design, deliver, and operate production-grade software, data, and AI systems. From first idea to long-term reliability in production.

Backend & Systems Engineering

Design and build reliable software systems and internal tools.

  • Backend APIs and services designed for long-term maintainability
  • Automation and internal tooling that reduce operational load
  • Cloud-native systems built with reliability and observability in mind
Explore Approach →
Applied Data Science & Decision Support

Turn complex data into actionable insights and decision support.

  • Exploratory analysis focused on actionable patterns and decision clarity
  • Statistical modelling and experimentation with explicit assumptions and limits
  • Decision metrics and performance analysis tied to real operational outcomes
Explore Approach →
ML & Data Engineering

Robust data and ML pipelines built for real operational constraints.

  • Data pipelines and feature layers designed for consistency, versioning, and scale
  • Training, inference, and monitoring workflows built for production constraints
  • Traceability and reliability mechanisms to understand, debug, and trust models over time
Explore Approach →
Production AI & ML Systems

End-to-end AI solutions combining ML, GenAI and software engineering.

  • Predictive ML and GenAI workflows designed for controlled behavior and reliability
  • Evaluation, guardrails, and monitoring to manage failure modes and drift
  • Integration of AI components into existing systems and business workflows
Explore Approach →

Experience

My experience at various companies

DataDome
Real-time bot & fraud detection at internet scale

Worked on large-scale backend systems for real-time bot and fraud detection, supporting high-volume traffic and strict reliability requirements. Owned critical components spanning behavioural signal ingestion, validation pipelines, and fraud evaluation logic, with a strong focus on correctness, performance, and operational robustness in production.

IBI Group
Data & analytics supporting large-scale infrastructure projects

Delivered data analysis and modelling work to support decision-making on complex infrastructure and urban systems projects. Worked closely with engineering and domain experts to translate requirements into analytical outputs, focusing on clarity, traceability, and practical impact rather than purely exploratory analysis.

NEC Japan
Applied machine learning for industrial and operational systems

Designed and implemented applied machine learning and analytical components within industrial and operational contexts. Contributed to data modelling, evaluation workflows, and system integration, ensuring analytical outputs were reliable, interpretable, and aligned with real-world constraints beyond experimentation.

Airbus Helicopters
Software automation in documentation-heavy industrial environments

Built internal automation tools and supported LEAN industrialisation initiatives in a highly regulated engineering environment. Combined software development with process optimisation, contributing to cost, time and defect reduction while operating within strict documentation and coordination constraints.

Indian Institute of Technology
Research-driven engineering and experimental systems

Engaged in research-oriented engineering work combining software development, experimentation, and system prototyping. Collaborated with academic teams to design, build, and evaluate experimental systems, strengthening foundations in analytical thinking, modelling, and disciplined experimentation.

Gemalto (UK)
Secure software systems and embedded foundations

Developed secure software and embedded system components in a security-critical environment. Worked on authentication flows, secure communications, and system integration, gaining early exposure to reliability, security constraints, and documentation-driven engineering practices.

Aix-Marseille Université
Engineering foundations and applied technical research

Built a strong engineering and analytical foundation through applied projects and technical research. Worked across software development, systems modelling, and experimental work, establishing the discipline and rigour later applied in industrial and production environments.

Skills

A systems-first skill set spanning software engineering, data, machine learning and cloud delivery.

Software Engineering

Backend services, interfaces, performance and dependable delivery practices.

PythonTypeScriptJavaScriptJavaScalaC/C++BashAPI designDistributed systemsConcurrency & asyncPerformance profilingDesign reviews & RFCs

Homelab

Personal R&D for reliable software, infrastructure and applied AI

I maintain a personal experimentation environment where I can test ideas end-to-end, not only the software itself, but also how it behaves once deployed and operated. This allows me to surface risks, trade-offs, and failure modes early, before they reach users or production environments.

It’s a controlled space to validate assumptions early, understand system behaviour over time, and reduce delivery risk before changes reach real users or production environments.

Physical servers (self-hosted)
5
Virtual machines
3
Active containers
120+
What I test
  • Software behaviour: Interfaces, integrations, data workflows, and how changes propagate across services.
  • Infrastructure behaviour: Deployment patterns, stability under load, failure recovery, and service dependencies.
  • Reliability: Failure modes, performance bottlenecks, alerting quality, and observability signals.
  • Applied AI: Practical limits of models, evaluation challenges, and integration trade-offs in real systems.
Why it matters

This practice helps me deliver solutions that remain dependable beyond the first demo.

By testing both the product and the way it runs in real conditions, I reduce surprises, improve maintainability, and make trade-offs explicit - speed vs. cost, simplicity vs. scalability, and performance vs. robustness.

In practice, this leads to fewer surprises in production, clearer trade-offs during design, and systems that age more gracefully over time.

Examples
  • What happens to response time when a new feature is introduced?
  • How does a service behave when a dependency becomes slow or unavailable?
  • How quickly are issues detected and diagnosed through monitoring?
  • Do data or model changes silently break downstream workflows?
  • How safely can a system be updated, rolled back, or recovered?

This environment complements my professional work by allowing me to validate ideas independently and continuously, outside of delivery timelines.

Business Impact

Decisions grounded in financial reality, not technology for its own sake

Principles
  • AI and analytics are financial decisions, not technology experiments.
  • The objective is measurable impact: lower operating cost, higher asset productivity, reduced operational and compliance risk, and stronger margin resilience.
  • Under cost pressure, volatility, and talent scarcity, data-driven systems are among the few levers that improve performance without increasing headcount or capital expenditure.
Typical impact areas

These are the types of outcomes I aim for when designing and delivering the systems described above.

Cost reduction
Operating costs
Failures (predictive maintenance)
Quality defects
Productivity
Operating margin impact
Through automation, decision support, and reduced rework
Risk control
Reduction in operational risk
Earlier detection & alerts
Working capital
Inventory levels
Forecast accuracy
ROI & payback

Let's connect

AI transforms how we work, not why we work. That stays human.

Yash Thevalil
AI & Machine Learning · Software Systems
Connect on LinkedIn

Open to discussions around applied AI, machine learning systems, and production-ready solutions.

© 2025 Yash Thevalil