FDE | Building with an Agentic Lens & First Principles
Bengaluru / New Delhi, India • Portfolio • LinkedIn • Email
Senior Full-Stack Engineer with 5+ years of experience building scalable web applications, real-time data platforms, and healthcare software. Currently applying first-principles engineering to Agentic AI—designing reliable agent harnesses, memory systems, and domain-specific SLMs.
- Agentic Harnesses: State-machine workflows, deterministic boundaries, and automated evaluations.
- Memory Systems: Structured context retention (working scratchpads, session history, and semantic search).
- Domain SLMs: Fine-tuning (LoRA/QLoRA) and evaluating small language models on private data.
- Autonom: Built multi-RPC failover for a Solana trading platform, cutting transaction errors by 70%.
- API3: Redesigned analytics data pipelines, reducing platform latency by 40%.
- QuickSwap: Optimized DEX trading flows, improving initial page load by 80%.
- Hackathons: Winner at ETH Istanbul (2023) • Finalist at ETH India (2024).
- SLM Fine-Tuning Pipeline: Enterprise SLM adaptation and evaluation framework for private data.
- Agent Memory Engine: Multi-tier context retrieval and pruning system for autonomous agents.
- Flow3: Event-driven workflow automation engine built with Temporal and Kafka.
- Autonom DEX: Real-time data indexing and streaming infrastructure on Solana.
TypeScript Node.js Python React Next.js Redis Kafka Temporal PostgreSQL Docker Solana