Product Manager at Centella AI. I work at the edge where AI research stops being impressive and starts being useful — building drug discovery tools with medicinal chemists, not just for them.
I’m a Product Manager at Centella AI in Hyderabad, working on an AI platform for small molecule drug discovery. The platform covers generative design, ADMET prediction, retrosynthesis and virtual screening — I own the product across all four areas.
I came into PM through engineering. CS background, Spring Boot and Python in production, system design, microservices — which means I can sit in an engineering review and understand the real tradeoffs, not just the slide deck version.
Early-stage startups ask you to do more than one thing. At Centella I also coordinate hiring and run social media. Before product: 2+ years as Campus Lead at EdVenture Park, running founder events and building a student startup community in Hyderabad. Good training for understanding why people don’t adopt things that seem obviously useful.
CS from VNRVJIET (9.39 CGPA), diploma from Govt. Polytechnic Hyderabad (9.9 CGPA). Currently open to conversations with PMs, founders, and researchers working on hard problems in AI and life sciences.
Owned the full product lifecycle for Centella’s generative design module — discovery research with medicinal chemists through to launch. Core challenge: making scientists trust novel structures they didn’t come up with themselves.
Roadmap and UX for ADMET prediction integrated with virtual screening. Surfacing model uncertainty in a way non-ML scientists can actually act on.
Generative Molecule Design — Product Manager, Apr 2024–present
Centella’s generative design module lets medicinal chemists specify target properties and receive novel candidate structures from the ML models. I owned the full lifecycle: discovery research, PRD, engineering prioritisation, and validation before each release.
The hardest problem was trust. Generative AI outputs look plausible and can be chemically wrong. I worked with the ML team to surface synthesisability scores and uncertainty estimates alongside structures, so chemists had something to apply their scepticism to.
Requirements and user research with synthetic chemists to bring retrosynthesis prediction into the platform. Hardest question: which constraints the AI had to honour.
Social media strategy, hiring coordination, internal operations at Centella. Campus Lead at EdVenture Park: entrepreneur events, Founders Friday, 2+ years building the student startup community across Hyderabad campuses.
ADMET & Virtual Screening — Product Manager, Apr 2024–present
ADMET prediction — absorption, distribution, metabolism, excretion, toxicity — filters poor drug candidates before wet-lab spend. I owned the roadmap. The design challenge: presenting confidence intervals in a way non-ML scientists could act on, not just file away.
Several rounds of research and iteration with scientists used to thinking in experimental error bars, not ML uncertainty. The vocabulary had to match theirs before the data could move their decisions.
Retrosynthesis Planning — Product Manager, Apr 2024–present
AI retrosynthesis works backwards from a target molecule to available starting materials. I defined product requirements and ran research sessions with synthetic chemists to understand which constraints the AI needed to enforce and which outputs they’d never accept without verification.
That research shaped both what the model was scoped to do and how routes were ranked in the interface.
Ops, Growth & Community — 2022–present
At an early-stage startup the PM role is rarely the only one. At Centella I coordinate hiring, run social media, and handle internal operations alongside the core product work.
As Campus Lead at EdVenture Park (2022–2024), I ran founder events and Founders Friday sessions across Hyderabad’s student ecosystem. Good training for understanding how people actually evaluate new ideas.
Product and UX for a peptide sequence design tool built for pharma clients. Bridging generative model outputs with wet-lab feasibility constraints.
Interface and pipeline tooling for molecular dynamics simulation runs. Reducing setup time and making simulation outputs interpretable without domain expertise.
Peptide Design Platform — Client Project
Pharma clients working on peptide therapeutics needed a way to explore sequence design space without running wet-lab iterations on every candidate. I led product definition and UX: from understanding how medicinal chemists think about sequence-activity relationships to specifying what the generative model needed to output for the results to be actionable.
Core challenge was the gap between AI-generated sequences and synthesisability. The platform had to surface constraints the chemist could act on, not just a ranked list of sequences with no context.
MD Simulation Workflow — Client Project
Molecular dynamics simulations are computationally expensive and hard to interpret without deep domain expertise. A client needed a workflow tool that let their team set up, run and interpret simulations without needing a dedicated computational chemistry team on every run.
I defined the product requirements and worked with engineers on the interface layer. Key decision: which parameters the UI should expose and which should be defaulted intelligently, so scientists spent their time on the science, not the setup.
Roadmap prioritisation, discovery research, PRDs, stakeholder alignment, OKR setting, GTM for technical AI products.
CS foundation: system design, networking, OS, microservices, backend. I understand engineering tradeoffs, not just the slide deck version.
Contextual inquiry, diary studies, expert user testing with scientists. Getting to the real problem before starting to solve it.
Generative design, ADMET, retrosynthesis, virtual screening. Working knowledge of how medicinal chemists actually think.
ML, model evaluation, data pipelines. Applied in drug discovery contexts. Comfortable with the tradeoffs in AI product decisions.
Growth, social media, hiring, internal ops. Full picture of how early-stage companies work — from the inside.
Building AI tools for scientists, researchers, or domain experts? I can bridge the technical and human sides of that challenge. Let’s talk.