SachinJain
Technical Product Manager shaping and building AI-native products and scalable systems
From enterprise platforms in financial services to AI-driven products, I lead product thinking and build systems that are structured, scalable, and usable.
A career shaped by systems, and product judgment.
UBS
Led product and platform development for a global swaps processing system, driving the transition from legacy architecture to scalable microservices while enhancing system resilience, operational efficiency, and client-facing capabilities.
Carnegie Mellon University
Six-month capstone project in collaboration with Informatica, focused on designing and building a scalable stream processing framework using Apache Storm, capable of handling high-volume real-time data across constrained environments.
Credit Suisse
Contributed to the development and evolution of trading applications for index and exchange-traded funds, working within service-oriented systems to deliver incremental enhancements aligned with business and regulatory requirements
Exploring AI Capabilities and How They Fit into Building Products
Hands-on work across retrieval, agents, and workflows to understand what AI can do and how it can be applied in real product experiences.
What I’ve used
What I’ve learned
What I doing next
From system thinking to product execution
The next scene moves from how I operate to what I build.
A collection of work, ideas, and experiments
Building to understand, evaluate, and apply ideas in real-world contexts. Testing ideas against real use, beyond theoretical potential.
Prep Room
Prep Room is an AI-powered interview preparation system designed to help users build clear, structured, and credible answers through guided workflows rather than one-off responses.
Instead of treating interviews as isolated questions, it organizes preparation into reusable building blocks such as highlights, experiences, and refined answers and uses them to support continuous improvement over time
Interview preparation is usually fragmented across notes, documents, and generic chat threads. That makes answers harder to refine and harder to reuse.
Users work through answers, save stronger material, and build a structured base of moments and playbooks that can be reused across interview scenarios.
Designed system architecture across backend, frontend, Data, and AI layers with a multi-step workflows.Implemented context-aware retrieval for personalization.
The product became a good test of how AI feels when it is embedded into a workflow instead of acting like a single detached chat box. Contract-based communication improves reliability between agents
Immersive Yatra
AI-powered road trip planner for building detailed multi-day travel plans with structure, flexibility, and practical routing in mind.
Designed to turn vague travel ideas into richly organized itineraries that feel practical but has room for personal touch
The challenge was to create something that organize routes, timing, and stops into a single flow but leave room for personalization.
The system takes trip intent and transforms it into structured, multi-day planning with more useful pacing, organization, and itinerary detail.
Defined a structured output format for itineraries (routes, days, stops, stays). Iterated on prompt design to generate consistent. Focused on sequencing and flow rather than isolated recommendations
AI outputs improve when the prompt aligns with intent rather than over-specifying behavior. Too much constraint reduces usefulness, while the right level of openness improves outcomes.
Personal Research Agent
A focused research assistant designed to synthesize insights across selected papers and material
A personal research agent that synthesizes information across documents (PDF, DOC, TXT), ensuring outputs are grounded in selected source material using RAG. It supports research writing and poster preparation by organizing insights into clear, usable summaries.
Research material is scattered across multiple papers and it is time-consuming to extract and connect key insights.
Documents are processed and stored in a FAISS vector database. Relevant sections are identified based on the user query. Retrieved content is processed to generate clear, grounded outputs
Built a custom RAG pipeline and indexed a FAISS-based vector database for efficient retrieval. Implemented a two-pass approach to extract relevant context along with source references for verification.
Chunking strategy and token usage must be balanced — too little context reduces quality, while too much increases cost without better results.
Skills, systems, and tools in my stack.
This section combines how you think with what you use, so it grows naturally as your work evolves.
A continuous view of learning
What I’m learning, in practice and in motion. A mix of certifications, ongoing study, and papers that shape how I think and build.
Certifications
Learning Themes
Read Archive
Open for the next build
See Something, Say Something.
Explore my work, or reach out to connect...
Let’s turn ideas into product.
Open to conversations about building products and systems, from enterprise platforms to modern AI-driven applications