Building agentic systems
that scales in production

Founded Dobr.AI to raise the standard of tech hiring: replace obsolete interview frameworks and evaluate real-world skills instead of brand pedigree.

• Built a voice-based AI interview agent for natural conversational technical interviews that adapt to candidate responses and task progress.
• Architected a cloud-native AWS microservices platform for thousands of concurrent interviews with >99.9% uptime.
• Built an in-house agent orchestration framework with context-driven sub-agents for adaptive response generation, evaluation, and feedback - caching and preemptive processing to hold ~1.5s latency at low cost.
• Shipped cheating-detection from behavioral signals (response cadence, fraud signals) and dynamic validation agents.
• Designed a plugin system for new interview formats (machine coding, system design) and evaluation logic without core refactors.

Voice AIAgentsAWSMicroservicesLatency

Technical leader of a 40+ person engineering org. Architecture and execution from feasibility and technical POC through pilot and production rollout - 90%+ on-time delivery and 70%+ pilot-to-production conversion.

• Led the shift from traditional software engineering to an agentic engineering lifecycle with Auto Product Engineering: autonomously develops features from product specs and continuously improves production systems from real data.
• Delivered a 5× engineering productivity gain without quality loss - three months of projects in two weeks.
• Built a high-accuracy Graph RAG knowledge platform for banking, with evaluation benchmarks for high-density query scenarios at 98%+ accuracy when combined with Auto Product Engineering.

Staff ICGraph RAGAgentic SDLCBanking AIOrg leadership

Tech Lead across GenAI Kindle, AWS self-checkout fraud & abuse, and Amazon GO ground-truthing / camera planning - architecture, planning, and teams of 4–5 SDEs.

• GenAI Kindle: first-generation GenAI-enabled Kindle workflows - robust model deployment, LLMs, agentic execution, and safeguards against harmful or biased content. Cross-team collaboration; led 5 SDEs.
• Fraud & abuse (AWS self-checkout / Fresh): end-to-end suite tackling retail shrinkage ($94.5B industry loss in 2021) - theft detection, alerts, MLOps, continuous training/deployment, ground-truthing. Led 4 SDEs.
• Amazon GO receipt ground-truthing: cut associate effort 50% and metric turnaround 66% (3 days → 1 day).
• Camera planning automation: streamlined planning 60% and reduced SDE operational effort 95%.

GenAIFraudMLOpsRetailKindleLeadership

• First engineer. Built and optimized mobile apps with modern architectural patterns.
• Redesigned backend message routing to cut latency by >30%.

MobileBackend0→1

Student-founded mobile software startup focused on Android UX and productivity utilities.

• Reached over 3 million users.
• Covered by Gizmodo, Lifehacker, and XDA.
• Recognized across platforms including IIT Madras and IIT Mandi, with global interest.
• Only student speaker at the Official Docker Meetup, Hyderabad.

AndroidConsumerGrowth