I ship agentic AI and SaaS products that businesses trust in production.
Senior Product Manager with 12+ years turning messy business workflows into AI and SaaS systems people actually rely on — intelligent document processing, LLM productization, and agentic in-product systems. I also design the hardware behind a small IoT venture on the side, which keeps my product judgment grounded in what's actually buildable.
About
I'm a Senior Product Manager who ships AI and SaaS products into production for real customers — intelligent document processing, LLM productization, and agentic in-product systems that businesses can actually trust with their workflows. That means owning the full loop: defining what "good" looks like with evals, deciding when an agent earns its complexity over a simpler workflow, and weighing accuracy against cost and latency as real product trade-offs, not afterthoughts.
I don't just write specs for engineering teams — I've built enough of the underlying systems myself, including a side hardware-plus-SaaS venture called Athlax Solutions, to know what a spec is usually missing. That grounding shows up as measurable efficiency gains from AI-native workflow changes at work, and as a growing set of SaaS products I'm building end to end, from PRD to prototype.
Formally: a B.Eng in Electronics & Communication, followed by a postgraduate program in AI & Machine Learning. Practically: whatever I've had to learn to get a product from idea to something that actually runs.
- ROLESenior PM, Procys IDP
- COMPANYOpenprovider
- VENTUREAthlax Solutions
- EDUCATIONB.Eng, Electronics & Communication — VTU
- EDUCATIONPGP, AI & ML — Great Learning
- LOCATIONAiroli, Navi Mumbai
Experience
Leading product for Procys, an Intelligent Document Processing platform. Own the roadmap across ERP integration and LLM migration, and built out Isabelle, a support chatbot now handling 500+ conversations a week. Drove AI-native workflow changes internally that produced measurable efficiency gains across the team.
A hardware-plus-SaaS venture spanning industrial IoT, smart safety hardware, and a small hospitality/productivity SaaS line. I run product, hardware design, and firmware across the portfolio — see Projects below for the individual bets.
How I build AI & SaaS products
- Define quality before building
- Every AI feature starts with an eval set and a shipping bar agreed with engineering and the business. If we can't measure it, we don't ship it.
- Design for the failure, not the demo
- I map how the model will be wrong, then decide what the product does about it: validation rules, confidence thresholds, or a person in the loop.
- Choose agents only when they earn it
- Many problems are better solved by a single well-built prompt or a deterministic workflow. Agents are for tasks that need planning and tool use.
- Weigh accuracy, cost, and latency together
- The best model is the one that meets the bar at a price and speed the business can sustain. I make those trade-offs explicit.
- Keep people accountable for outcomes
- AI can draft, extract, and recommend. Decisions with real business weight keep a named human owner and an audit trail.
- Ship SaaS end to end
- From PRD to prototype to go-to-market — I scope products I can actually validate with users before asking an engineering team to build them.
Projects
Hospitality SaaS combining property management, energy management, and inventory control in one platform.
Shipped so far: full PRD and an interactive prototype. Currently extending with an Inventory Management module — room amenities, housekeeping supplies, and engineering parts, with presence-detection-triggered restocking alerts.
An AI tool that reads legal documents and surfaces the parts that matter — clauses, obligations, deadlines, risks, and penalties — without a line-by-line manual read.
Includes: a natural-language Q&A interface that answers strictly from the
document's own content.
Target: law firms and legal professionals, launching into Tier-1/2 Indian firms
and corporate legal teams first. End-to-end PRD complete.
A personal productivity app built around a "timestamp everywhere" philosophy — every action carries a record of when it happened.
Stack: React + Supabase. PRD and MVP scope defined.
A financial planning calculator modeling combined systematic investment and systematic withdrawal plans, built for potential public release.
Industrial fire safety compliance IoT platform under Athlax (formerly named Sentinel). Go-to-market is pivoting toward FLA and fire-AMC contractors, selling white-label compliance evidence dashboards.
Stack: ESP32-S3, 4G LTE, LiFePO4 power chain, ATECC608B secure element,
ThingsBoard/TimescaleDB backend, remote monitoring over MQTT with device pairing and
multi-device dashboards.
Positioning: built around CPCB fire compliance requirements — one of two primary
product bets in the Athlax hardware+SaaS portfolio.
A smart electrical spike guard, engineered for the Indian safety-certification market.
Status: working toward BIS IS 302 certification.
The other of the two primary Athlax hardware bets, alongside FARMS.
A series of ESP32-based systems for remote field monitoring and control, designed and built end to end — schematic, firmware, cloud, dashboard.
Includes: FSDS (Field Status & Diagnostics System) over WiFi/4G with Firebase RTDB and a custom dashboard; an ESP32 solenoid valve controller with deep-sleep power management; a 22-node ESP-NOW solenoid mesh for field deployment; an ESP-MESH train monitoring architecture; and JalMitra, a solar-powered animal water dispenser (PRD and BOM complete).
Skills
PRODUCT & AI
HARDWARE & EMBEDDED
Writing
I write regularly about intelligent document processing, applied LLM systems, agentic products, and the business of building SaaS — drawn from the work, not from the outside. Around 100 essays so far; a few to start with:
Why OCR Accuracy Is the Wrong Metric — Use Straight-Through Processing Rate Instead
OCR accuracy is a component metric that tells you almost nothing about whether your document automation product is working. Straight-through processing rate does.
Read the full piece →Hybrid Routing Between Cheap Models, Expensive Models, and Humans
A three-tier routing architecture and the hard part nobody talks about: producing a difficulty signal good enough to route on.
Read the full piece →Lessons From Running a Customer-Facing Assistant at 500+ Conversations a Week
What changes when an in-product assistant stops being a demo and starts carrying real conversation volume, week after week.
Read the full piece →Certification as a Moat: The Unglamorous Regtech Advantage
Certifications, accreditations, and registrations are slow, expensive, and boring — which is precisely what makes them defensible in a market where software features are not.
Read the full piece →Hardware Plus SaaS: The Hardest Business Model, and the Most Defensible
Why connected-hardware businesses are punishing to operate and extremely difficult to displace once established — and what has to be true for the trade to be worth it.
Read the full piece →Twelve Years of Shipping: The Mistakes I Keep Repeating
The errors that recur despite being known, why knowing about them is insufficient, and the structural fixes that actually work.
Read the full piece →Contact
Hiring for an AI PM or FDE role? Let's talk.