anurag sikdar · ai product owner

// 01 · who i am

about me.

ai product owner with 3+ years at the intersection of generative ai engineering and product delivery — a builder-turned-owner who ships.

at demand nexus, took three products from zero to production in the first 60 days: an internal hrms, an ai-assisted b2b crm, and lara, a live speech accent coach. now leading project x, a stealth ai initiative for the hospitality industry.

working fluency with frontier models — anthropic claude 5, openai gpt-5, google gemini 3 — enables realistic scoping, sharp build-vs-buy and model-selection calls, and eval-driven roadmaps that engineering teams trust.

proven record writing prds and testable prompt specs, running agile delivery end to end, aligning cross-functional and client teams, mentoring engineers, and contributing to pre-sales pursuits across saas, fintech, loyalty, and industrial domains.

↳ the one who actually ships it.

// 02 · what i bring

skills & tools.

product ownership

product roadmapping discovery & research prds & user stories acceptance criteria backlog management rice / moscow okrs & kpi design a/b experimentation gtm planning stakeholder mgmt pre-sales

agile & pm tools

scrum sprint planning retros effort estimation jira confluence linear notion clickup miro trello

product design

open design figma wireframing prototypes ux flows design systems usability testing

frontier models

claude 5 (fable / mythos) opus 4.8 / sonnet 4.6 gpt-5 / gpt-4 gemini 3 / 2.5 llama 3 mistral qwen deepseek

ai systems

rag pipelines agentic workflows react langchain / langgraph crewai mcp human-in-the-loop llm evals (ragas, trulens) fine-tuning (lora, qlora) faiss / pinecone / weaviate neo4j voice ai (stt/tts) semantic search

engineering fluency

python sql javascript fastapi flask docker kubernetes ci/cd (github actions) aws / azure / gcp rest / grpc postgres / mongo / redis

// 03 · where i've shipped

experience.

ai product owner

demand nexus · pune, india

may 2026 – present

own the end-to-end lifecycle of the company's ai product portfolio — discovery, prds, backlog, sprint execution, launch — shipping 3 production products within the first 2 months.
shipped an internal hrms digitizing the core employee lifecycle (onboarding, attendance, leave, approvals), replacing spreadsheet-driven hr operations with a single self-serve system.
launched an ai-assisted b2b crm covering the full lead lifecycle — capture, enrichment, scoring, and pipeline tracking — giving sales one source of truth for b2b lead management.
shipped lara, a live speech accent coach delivering real-time pronunciation and fluency feedback over a streaming stt–llm–tts pipeline; defined the scoring rubric, latency budgets, and corrective-feedback ux.
leading project x (stealth) — an ai product for the hospitality industry — owning market discovery, mvp scoping, feasibility validation on frontier models, and go-to-market planning.
run product rituals end to end (sprint planning, backlog grooming, retros, stakeholder demos) and prioritize with rice — keeping engineering, design, and business on one roadmap.
drive per-feature model strategy across claude fable 5, gpt-5, and gemini — trading off quality, latency, and cost — and de-risk every build with open design / figma wireframes and clickable prototypes before a line of code is written.

gen ai engineer & llm systems architect

botmatic solutions pvt ltd · pune, india

aug 2025 – may 2026

delivered large-scale gen ai capabilities for myvyay 3.0, an enterprise saas platform: multi-step agentic pipelines for travel planning, compliance enforcement, invoice ocr, and whatsapp conversational interfaces serving enterprise customers across domains.
built a cross-product prompt framework standardizing agent behavior across gpt-4/5, claude, qwen, and open-source models — cutting prompt duplication by 60% and accelerating iteration across product lines.
engineered production voice agents on gemini 2.5 multimodal with sub-second latency, reducing human agent workload by ~40%.
ran systematic a/b tests, prompt evaluations, and user simulations to raise accuracy and cut hallucination — reducing policy violations by 85% and manual processing by ~70%.
partnered with product managers, engineers, and ux designers in agile sprints to translate saas workflows into agentic ai tasks; containerized and deployed gen ai microservices on gcp via docker/kubernetes with rest integrations into enterprise systems.

ai automation engineer — fintech

claimmyshares · pune, india

feb 2025 – aug 2025

designed agent-driven automation for client tracing and recovery in regulated fintech, replacing manual kyc workflows with llm-powered documentation systems that improved tracing accuracy and compliance.
built pipelines for dynamic report generation and natural-language query interpretation over sensitive financial data, ensuring accuracy, auditability, and regulatory alignment.
translated complex regulatory processes into structured agent logic, enabling cross-functional teams to run compliance workflows independently — without engineering support.

ai product developer & analytics engineer

paytc loyalty · pune, india (hybrid)

nov 2023 – jan 2025

built kpi frameworks and llm-powered analytics pipelines that gave product leadership actionable insight, directly driving data-informed decisions across the loyalty saas platform.
automated reporting with llm summarization and embedding-based semantic search across crm and loyalty data, reducing manual analytics overhead.
partnered with cross-functional product teams to define ai use cases and turn business requirements into testable prompt specifications and agent workflows.

ai research analyst — predictive & industrial nlp

bharat forge ltd (trainee) · pune, india

mar 2023 – nov 2023

built predictive maintenance models on real sensor data using scikit-learn and classical ml — statistical analysis, anomaly detection, and nlp-driven insight extraction under production constraints.
automated maintenance documentation pipelines, translating model outputs into operational systems and improving equipment reliability for one of india's largest industrial manufacturers.

// 04 · what i've built

key projects.

framework

benchmarking & quality

multi-model eval

cross-model system benchmarking agent performance across gpt, claude, gemini, and deepseek — automated a/b comparisons, hallucination detection, and output quality scoring to inform optimization at scale.

benchmark everything.

desktop

desktop ai agent (electron)

kira

production-grade desktop agent for browser automation, spreadsheet management, email handling, and system task execution via natural-language commands with multi-turn context.

your desktop, automated.

voice ai

real-time voice pipeline

raya

end-to-end voice pipeline (stt, llm reasoning, tts) for task management, scheduling, and contextual conversation — the groundwork later applied to shipping lara.

talk to your ai.