Headline

Member of Technical Staff @ Aegion | AI Agents, LLM Pipelines & Backend Infrastructure | Python · TypeScript · Kafka · Kubernetes | Winner ETH India ’24

About

I build AI systems that have to survive production, not just demos. At Aegion (Mira), I work on a small team shipping AI agents for real businesses: a customer-success agent with human-in-the-loop review — every drafted reply gets accepted, edited, or rejected, and those outcomes are instrumented as the quality metric — plus call transcription and LLM scoring pipelines, and the analytics platform underneath (Kafka, Postgres, Kubernetes). Earlier at Mira I worked on LLM API infrastructure handling 2M+ requests a month: routing, auth, rate limiting, billing. The through-line: I like owning a loosely-defined problem end to end, schema to deploy. And I care about the unglamorous parts — observability, feature flags, contract checks at pipeline boundaries — because that's what makes AI systems trustworthy. Side project: Sentinel, a self-healing dev system where an LLM turns GitHub issues into 100+ reviewed PRs a month, with MCP-based production log analysis. Winner, ETH India 2024 (4 track prizes) and HackQuest 2024. B.Tech CSE, IIIT Sri City. Stack: Python (FastAPI), TypeScript (Node.js, React), Kafka, PostgreSQL, Redis, Kubernetes, AWS.

Location

Bengaluru

AI stack

claude code

gemini

langchain

langfuse

ollama

opencode

Work experience

Member of Technical Staff at Aegion AI, Mar 2026 - Present · 6 mos, Bengaluru, Karnataka, India, Building aegion.so — AI agents and analytics for real businesses. • Own a full-stack analytics product end to end: CRM data sync, Postgres warehouse (Drizzle ORM), call transcription + LLM scoring pipelines, and React dashboards — shipped in two-week cycles and used daily by clients. • Run a production customer-success AI agent with human-in-the-loop review (accept/edit/reject outcomes instrumented as the quality metric) and RAG over Milvus. • Delivery tooling: database-driven feature flags for deploy-free rollouts, New Relic observability; cut a key analytics endpoint from 71s to 2s.

Software Development Engineer at Mira, Apr 2025 - Present · 1 yr 5 mos, Singapore, • Architected LLM API infrastructure handling 2M+ requests/month: routing, authentication, rate limiting, and billing approval workflows for 50+ enterprise clients (40% revenue growth). • Built a product-analytics platform end to end: Kafka producer/consumer services with multi-worker scaling, batched OpenSearch writes, and unified Python/JS SDKs adopted by two consumer AI products. • Ran microservices processing 10,000+ transactions daily over Kafka and BullMQ at 99.9% uptime, with New Relic tracing and alerting across API endpoints.

SDE Intern at Mira, Jul 2024 - Apr 2025 · 10 mos, Remote

Data Analyst at Centre for Railway Information Systems (CRIS), May 2024 - Jul 2024 · 3 mos, Remote, Developed an advanced machine learning model for predicting passenger train waitlists. Conducted a comprehensive literature review, preprocessed large datasets, and applied clustering techniques like K-Means, Hierarchical Clustering, and DBSCAN to uncover meaningful patterns in passenger PNR data. This experience enhanced my skills in data analytics, machine learning, and problem-solving within the transportation sector.

Full Stack Developer at Zymo, Mar 2023 - Mar 2023 · 1 mo, Remote

Back End Developer at Propertybhavan, Nov 2022 - Jan 2023 · 3 mos

Education

Bachelor of Engineering - BE, Computer Science, Indian Institute of Information Technology, SriCity

12th, Studying, Allen

Public links

github: https://github.com/dhrupad-sah