AI Engineer building production-grade agentic systems and enterprise AI platforms.
Internships, hackathons, and the systems in between.
I’m an AI Engineer building production agentic systems and enterprise AI platforms. Shipped applications across logistics, e-commerce, customer support, CRM, QA, and voice — with depth in multi-agent orchestration, enterprise integrations, evaluation, and observability.
The work spans agentic customer support, voice AI, simulation engines, and internal automations — QA and business-team tooling — on top of LLM infrastructure. Currently looking for opportunities where I can design, build, and scale AI systems end to end.
B.Tech (Honours), Artificial Intelligence — CGPA 9.0/10
2023 — 2027
Senior Secondary (XII), Science — 93%; Secondary (X) — 93%
2018 — 2022The stack I use to design, build, and run AI systems.
My professional journey and internships in AI, ML, and Software Engineering.
AI Developer Intern
•Dec 2025 — Aug 2026
•Houston, TX, United States (Remote)
Built and shipped AI agents for logistics & e-commerce clients across Gmail, Outlook, and WhatsApp — auto-classifying inbound emails, drafting responses, and triggering ETA follow-ups — plus a real-time voice AI system (LiveKit, Twilio) for call routing and escalation, cutting manual support effort by ~60 hrs/week and improving response time by ~90%. Architected multiple multi-tenant AI agent systems in Python/FastAPI, enabling new client workflows to ship by configuration instead of rebuild. Built an AI-first internal operating platform (Next.js, React, TypeScript) spanning 18+ enterprise portals on MongoDB + Prisma, improving cross-functional efficiency ~75%. Authored architecture and API documentation adopted across the team; onboarded new engineers and demoed AI workflows to non-technical stakeholders.
Cloud Engineering Intern
•May 2025 — Jun 2025
•Pune, Maharashtra, India (Remote)
Deployed Terraform-based infrastructure-as-code to automate AWS resource provisioning and management across client environments, achieving 50% faster deployment cycles. Cut cloud costs by 60% by consolidating redundant repositories, implementing scale-to-zero for SIT containers, and migrating scheduled GitHub Actions workflows to AWS Lambda + EventBridge.
President (MATLAB and Open-Source Club)
•Aug 2024 — Feb 2026
•Bhilai, Chhattisgarh
Led the MATLAB and Open-Source Club, organizing workshops and events around AI, data science, and open-source tools. Mentored peers on projects spanning machine learning, automation, and engineering fundamentals. Built a culture of experimentation and sharing within the student developer community.
Real-time policy engine in Rust that intercepts every AI agent tool call and returns ALLOW / BLOCK / ESCALATE in under 6ms with zero LLM dependency in the decision path. Features a tamper-evident append-only audit ledger with cryptographic checkpoints, mapping to OWASP Agentic Top 10, EU AI Act, and SOC 2 compliance frameworks.
Unified LLM gateway delivering one OpenAI-compatible API across any provider, with load balancing, cross-provider failover, circuit breaking, rate limiting, response caching, and content guardrails. Sustains 10K–70K req/s at 0.5–1ms added latency in under 7K lines of Go, using 1–2% CPU and 50–150MB RAM under burst load.
Agentic simulation platform that turns seed data into interactive digital worlds for scenario planning. Memory-driven agents (CAMEL-AI) run a 4-stage pipeline — graph build, environment setup, simulation, reporting — grounded in a structured knowledge model (OWL/RDF) on Neo4j with vector-based recall, automated quality scoring, and an interactive network-graph dashboard.
Led a 5-member AI team at Smart India Hackathon 2025 to build a space-mission AI companion for astronauts. Engineered the multimodal inference pipeline — training CNN-based facial emotion recognition and speech models in TensorFlow/Keras and integrating them with conversational AI — achieving 92% emotion-detection accuracy at sub-200 ms latency.
LangGraph-based Gmail automation system that triages emails, drafts contextual replies, and schedules meetings with 100% human-in-the-loop approval. Architected an interrupt-driven workflow so users can review and override actions before anything is sent. Designed a persistent memory layer that learns user preferences across triage rules, tone, and scheduling to continuously improve personalization.
Built for NASA International Space Apps Challenge 2025 as a multi-agent, multimodal RAG engine over 600+ space-biology papers. Achieved ~98% citation accuracy and ~80% faster research insights compared to manual search. Designed agent-based retrieval and routing so the system can choose the right tools and context for each query.
Designed a multi-agent autonomous trading system using MCP servers for real-time market data. Achieved ~15% higher returns than a baseline RL setup across 50+ simulated trading sessions. Architected a 4-agent collaborative framework for strategy generation, evaluation, execution, and risk control with decision latencies under 200ms.
Swipe to explore →
Real-time policy engine in Rust that intercepts every AI agent tool call and returns ALLOW / BLOCK / ESCALATE in under 6ms with zero LLM dependency in the decision path. Features a tamper-evident append-only audit ledger with cryptographic checkpoints, mapping to OWASP Agentic Top 10, EU AI Act, and SOC 2 compliance frameworks.
Unified LLM gateway delivering one OpenAI-compatible API across any provider, with load balancing, cross-provider failover, circuit breaking, rate limiting, response caching, and content guardrails. Sustains 10K–70K req/s at 0.5–1ms added latency in under 7K lines of Go, using 1–2% CPU and 50–150MB RAM under burst load.
Agentic simulation platform that turns seed data into interactive digital worlds for scenario planning. Memory-driven agents (CAMEL-AI) run a 4-stage pipeline — graph build, environment setup, simulation, reporting — grounded in a structured knowledge model (OWL/RDF) on Neo4j with vector-based recall, automated quality scoring, and an interactive network-graph dashboard.
Led a 5-member AI team at Smart India Hackathon 2025 to build a space-mission AI companion for astronauts. Engineered the multimodal inference pipeline — training CNN-based facial emotion recognition and speech models in TensorFlow/Keras and integrating them with conversational AI — achieving 92% emotion-detection accuracy at sub-200 ms latency.
LangGraph-based Gmail automation system that triages emails, drafts contextual replies, and schedules meetings with 100% human-in-the-loop approval. Architected an interrupt-driven workflow so users can review and override actions before anything is sent. Designed a persistent memory layer that learns user preferences across triage rules, tone, and scheduling to continuously improve personalization.
Built for NASA International Space Apps Challenge 2025 as a multi-agent, multimodal RAG engine over 600+ space-biology papers. Achieved ~98% citation accuracy and ~80% faster research insights compared to manual search. Designed agent-based retrieval and routing so the system can choose the right tools and context for each query.
Designed a multi-agent autonomous trading system using MCP servers for real-time market data. Achieved ~15% higher returns than a baseline RL setup across 50+ simulated trading sessions. Architected a 4-agent collaborative framework for strategy generation, evaluation, execution, and risk control with decision latencies under 200ms.
National and global hackathons — where ideas become working systems in 48 hours.
Built BioSage, an AI-powered space biology knowledge engine that ingests 600+ NASA and research papers. Delivered ~98% citation accuracy with a multimodal RAG pipeline and agent-based retrieval. Enabled researchers to pull insights ~80% faster than manual search.
Led a 5-member AI crew to develop MAITRI, an intelligent psychological companion for astronauts. Achieved 92% multimodal emotion detection accuracy with sub-200ms latency and strong stress-prediction performance. Focused on supporting mental well-being for space crews in high-stakes environments.
Looking for AI engineers who ship? I'm currently open to opportunities where I can design, build, and scale AI systems end to end. I will ignore all soliciting.
