Product strategy
From customer discovery to a roadmap the business can commit to.
- — Customer and requirement discovery
- — BRDs and PRDs
- — Prioritisation and roadmap shaping
- — 0-to-1 product development
The cover story
I build AI products that hold up in real workflows.
Product and technology leader working across strategy, architecture, and applied AI — taking ideas from first experiment to enterprise-scale production.
13+ years across software delivery, technical training, data and AI consulting, and product leadership. Since 2019 I have led product and engineering for AI hiring systems: skills assessments, candidate screening, and conversational interviews. The work runs end to end: requirement discovery and PRDs, validating ideas as proofs of concept, setting architecture direction, planning releases, and rolling out with enterprise clients.
Most of that production depth is in hiring technology. The patterns travel well beyond it — evaluating open-ended work, extracting structure from documents, orchestrating conversations, and keeping people in the loop where judgement matters.
01 · Leadership
One role, four disciplines
Architecture
Direction for AI systems that have to survive production load.
- — Architecture direction
- — RAG pipelines and LLM orchestration
- — Agent-based automation
- — Cloud microservices on AWS
Enterprise delivery
Turning demos into rollouts, and rollouts into adoption.
- — Release planning
- — Requirement workshops and demos
- — Client-facing rollout
- — Escalation handling
Teams
Building the people and operating rhythm behind the product.
- — Helped build and scale the India operation
- — Team building and mentoring
- — Delivery governance
- — Cross-functional coordination
02 · The systems
Three systems in production
01
Enterprise Skills Assessment Platform
Skills evaluation for enterprise hiring teams: coding and data-science tests, source-code verification, remote proctoring, and analytics across a 500+ skill library.
How it flows
Skill library → Candidate attempt → Proctoring signals → Code verification → Analytics
02
AI-Assisted Candidate Screening
A recruiter-facing assistant combining embedding search and LLM ranking for JD–CV matching, context-aware question generation, and structured candidate reports.
How it flows
Résumés + JD → Parsing → Embedding search → LLM ranking → Questions + report
03
Conversational Interviewing System
An agentic interview system running multi-turn conversations with dynamic follow-ups, live coding support, identity controls, and LLM-generated insights for reviewers.
How it flows
Candidate answer → Agentic interviewer → Follow-up or code task → Signals for review → Reviewer insights
03 · Architecture
The stack, layer by layer
| Layer | What it covers | Tools |
|---|---|---|
| Experience | Recruiter assistant · Assessments · Conversational interviews · Reporting | Product surfaces |
| Orchestration | Agentic workflows · LLM orchestration · Multi-turn flows | LangChain, Agents |
| Intelligence | LLMs · NLP · Prompt engineering · Classification | OpenAI APIs, Hugging Face |
| Retrieval & data | Embeddings · Vector search · Résumé parsing · Analytics | PostgreSQL, SQL |
| Platform | Microservices · REST APIs · Containers · Serverless | AWS, Docker |
04 · Career
A career in five chapters
I.
2019 — Now
Product & engineering leadership · Head of Product
Otomeyt AI, Bengaluru
AI hiring technology. End-to-end product thinking with hands-on AI system design. Led productisation of AI across résumé parsing, NLP, video analytics, and code evaluation. Delivered 7+ production AI systems and 15+ proofs of concept, converting several into enterprise-scale features. Helped build and scale the India operation with leadership.
II.
2017 — 2018
Big Data & AI Consultant
Paness IIHT, Cameroon
AI/ML consulting: chatbots, data mining, and analytics-led business support for clients. Translated business requirements into implementable technical solutions. Trained junior analysts; supported client delivery.
III.
2014 — 2017
Technical Manager
ASIT
Helped initiate a software-development division and managed delivery of new initiatives. Defined design and functional scope for a job-portal product. Trained employees and freshers on current technologies.
IV.
2013 — 2014
Independent Developer & Trainer
Self-employed
Built college-level software projects and trained students on development practices.
V.
2012 — 2013
Technical Trainer
NIIT
Technical training for freshers and working professionals.
Education: B.Com. (Hons.) in Accounting, LNT College.
05 · By the numbers
13+
Years across technology and product
7+
Production AI systems delivered
15+
Proofs of concept and R&D
2019
Leading product and engineering since
PoC to production
Converted multiple proofs of concept into enterprise-scale features used in high-volume hiring workflows.
AI-native platform
Embedded agentic workflows and LLM decision layers across screening, assessments, and interviews.
Scaled an operation
Helped build and scale the India operation, partnering on product direction and execution priorities.
Started a division
Helped initiate a software-development division and delivered its first initiatives.
“Accuracy alone doesn’t drive adoption. Workflow fit does.”
06 · The notebook
Fifteen-plus proofs of concept
An index of the areas explored. Several became production capabilities; a few remain research.
01Résumé intelligenceProduction
Structured extraction from unstructured résumés.
02Semantic JD–CV matchingProduction
Embeddings and vector search over role and candidate context.
03LLM ranking and shortlist reasoningProduction
Ranking candidates with explainable reasoning.
04Context-aware question generationProduction
Technical questions grounded in the role and the candidate.
05Structured candidate reportsProduction
Recruiter-ready summaries from screening signals.
06Source-code analysis and evaluationProduction
Evaluating submitted code beyond pass/fail tests.
07Video analytics for assessmentsProduction
Proctoring signals that support human review.
08Agentic interview orchestrationProduction
Multi-turn flows with adaptive follow-ups.
09Written and spoken answer evaluationDeveloped
Speech-to-text combined with LLM evaluation.
10Real-time voice turn latencyResearch
How turn-taking and latency shape a voice interview.
11Enterprise chatbotsClient delivery
Delivered for consulting clients, 2017–18.
12Data mining and decision supportClient delivery
Analytics-led business support for clients.
Principles
1. Workflow fit before model cleverness
An AI feature can win the demo and still fail in daily use. I judge it by whether it makes someone’s work meaningfully easier.
2. Prove it small, then productise
Validate as a proof of concept before committing a roadmap; carry what holds up into production.
3. Platform judgement
Know when to build a reusable capability and when to deliver a targeted customisation for an enterprise account.
07 · Further reading