Headline

Building AI analytics @Bicycle.ai | Agent Harnesses · LLMs in Production | Ex-upGrad | BITS Pilani

About

The hard part of an AI product isn't getting a model to produce something. It's getting someone to trust it enough to act on it. That gap is most of my job. At Bicycle.ai, I work on an AI data analyst copilot for revenue teams in e-commerce, travel, fintech, and logistics. It plugs into a company's existing data stack, watches business performance, catches anomalies early, and explains which drivers caused them. Priceline and Bigbasket are among the companies using it. I work both sides of the product. On the backend, the Python services and prompts that generate reports, dashboards, and KPI trees. On the frontend, the surfaces where someone reads what the AI produced, disagrees with part of it, edits it, and ships it anyway. I've owned our report authoring product for over a year, built the incident triage and anomaly training experiences from scratch, and contributed to the onboarding flow that took new-user setup from several days to under twenty minutes. The lesson that stuck: AI features fail differently from normal features. They fail plausibly. A broken API call shows you an error, but a confidently wrong insight shows you a chart. So a lot of the engineering that matters is unglamorous, making failure visible and reversible: explicit no-data states instead of blank screens, rollback when a pattern edit breaks, quality checks before generated output reaches a customer. I'm most useful where the problem is still vague, the model output isn't reliable yet, and someone has to ship it anyway. Working with: React, TypeScript, Python, FastAPI, LangChain and whatever's needed

Location

Greater Bengaluru Area

AI stack

claude code

codex

cursor

Work experience

AI Engineer at Bicycle AI, Sep 2025 - Present · 1 yr, Bengaluru, Bicycle is an AI-powered data analyst copilot that delivers continuous, actionable insights for high-volume transactional industries including e-commerce, travel, fintech and logistics. It plugs directly into existing data stacks and uses LLM and AutoML capabilities to start surfacing insights within minutes: identifying what changed, evaluating the key drivers, explaining why it happened, and recommending actions worth taking. Series A, funded by Decibel, Greylock, Lightspeed, Menlo and Fin Capital. Customers include Priceline.com, The Parking Spot, ACERTUS, UrbanPiper and Bigbasket.

Software Engineer at Bicycle AI, Jun 2025 - Aug 2025 · 3 mos, Hyderabad, Telangana, India

Strategist at upGrad, Aug 2024 - Dec 2024 · 5 mos, Bengaluru, Built an AI tool in React and FastAPI that generated tailored sales pitches from upGrad's program data, improving sales efficiency and conversion by 30%. Received a pre-placement offer (PPO) at the end of the internship

Education

Bachelor of Engineering - BE, Birla Institute of Technology and Science, Pilani

Public links

github: https://github.com/Nyyxoyy?tab=repositories