Reasoning infrastructure2026Active
Smriti
Version control for reasoning, with checkpoints, branching, and clean state restoration across long-running conversations and model switches.
Trying to understand intelligence by building it.
AI systems engineer working across production ML, intelligent systems, and applied research.
I build production AI systems, lead engineering teams, and spend most of my time where model quality meets compute, latency, reliability, and deployment constraints.
I keep returning to questions about intelligence, agency, memory, and what only becomes clear once ideas are forced into working systems.
12+ yearsproduction AI and software systems
15-20 peopleteams led across architecture and delivery
Idea to deploymentfrom early concepts to deployed systems
Currently exploring
- Reasoning systems, memory, and agent workflows.
- Production ML systems built for latency, reliability, and deployment constraints.
- Projects where research questions become usable software.
Selected work
Selected work.
Projects that best reflect the problems, abstractions, and systems questions I keep returning to.
Trust kernel for agents2026Active
Sentinel OS
A local-first trust control plane for agentic systems focused on policy enforcement, verifiable execution, and cryptographic proof of what actually happened.
Decision support system2026Prototype
PropOps
An AI-assisted property transparency tool that pulls together fragmented registry and legal signals to help buyers reason about risk before they commit.
Selected systems work2019 - PresentOngoing
Production AI Systems for Mobility
A confidentiality-safe snapshot of the kind of real-world AI systems work I have spent years doing: perception, in-cabin intelligence, optimization, and deployment under hard constraints.
Interests
Research interests.
Themes that keep showing up across both the systems I build and the questions I keep thinking about.
Intelligent systemsTrustworthy agentsMemory and stateRepresentation learningReasoning infrastructureReal-world ML deploymentSystems that remain legible under constraints
I am most interested in intelligent systems that remain legible under real constraints, especially around reasoning, memory, representation, and trust.
Experience
Experience.
Production AI, technical leadership, and systems built under hard constraints.
Sr. Tech Lead, AI/ML Autonomous Systems · KPIT Technologies
Leading and building production AI systems across mobility, perception, and in-cabin intelligence while staying close to architecture, optimization, and deployment.
- Led perception and in-cabin AI work across global mobility programs while staying close to implementation detail.
- Managed 15-20 member teams while driving architecture, execution planning, technical reviews, and delivery.
- Built multi-model, multi-sensor pipelines under hard latency, compute, throughput, and safety constraints.
Software Engineer · Independent / Freelance Work
Built product-oriented systems across mobile ML and full-stack application development.
- Developed a deep learning based neural style transfer system for mobile applications.
- Built a full-stack platform for healthcare billing workflows.
Software Engineer · Accenture
Worked on backend platforms, automation, and telecom systems where correctness, process, and modernization mattered.
- Built automated testing systems for telecom services.
- Contributed to rating, billing, and backend modernization work in production settings.
Philosophy
Working principles.
A few principles that shape how I approach both engineering work and longer-horizon research questions.
- Stay close to first principlesThe work should still make sense after you strip away tooling, hype, and surface detail.
- Let implementation sharpen the ideaBuilding is part of thinking. Implementation often reveals the real question faster than abstraction does.
- Prefer durable systems over polished demosA good prototype matters, but I trust work more when it survives constraints, complexity, and repeated contact with reality.
Building to understand. Understanding to build.
I like the loop between inquiry and implementation. Ideas become more precise when they survive contact with code, interfaces, data, and time.
Updates
Recent updates.
A short record of public releases, milestones, and notable changes.
PropOps public prototype released.
An early public prototype for AI-assisted property diligence, focused on making fragmented legal and registry risk easier to read.
Smriti public repo and demo released.
A reasoning-state system built around checkpoints, branching, and cleaner recovery from context drift across models and sessions.
MLOps Engineering 101 published.
A practical curriculum focused on turning ML knowledge into engineering systems teams can actually run and maintain.
Sentinel OS public milestone.
The repository captures a clearer direction for local-first policy enforcement and verifiable agent execution.
Links
Contact.
The projects and repositories are the best place to start. The links below are the easiest way to reach me or follow current work.
Reach out or follow along
Email is the easiest way to reach me. You can also find current public work through GitHub, LinkedIn, X, and Instagram.
Education
- M.S. Computer Science
University of Colorado Boulder
2024 - Present - Executive PG Programme in ML & AI
IIIT Bangalore
2022 - 2023 - B.E. Computer Engineering
Nagpur University
2013