AI & Software Engineer specializing in real-time voice & vision agents, high-concurrency backend pipelines, and intelligent multi-modal systems.
I'm Ishaan Aggrawal, an AI Engineer based in India. I build production-grade AI agents, low-latency voice pipelines, and scalable backend architectures.
"First, solve the problem. Then, write the code."
— John Johnson
Artizence Systems LLP • Remote
Multi-node conversational agent workflows with LangChain, LangGraph, ChromaDB, and evaluation harnesses.
Low-latency WebSockets & WebRTC token-pipelined voice agents using Whisper, Deepgram, and custom TTS.
YOLOv11 & ByteTrack real-time multi-camera tracking integrated with Vision LLMs for threat detection.
High-throughput FastAPI, Express, BullMQ job queues, Redis caching, and microservices architecture.
AWS Bedrock, Groq Llama-3.3-70B sub-200ms inference, parallel parsing, and prompt engineering.
Next.js, React, TypeScript, Tailwind CSS, PostgreSQL, MongoDB, Prisma ORM, and Supabase.
Replaced naive HTTP pipeline (~2-4s) with a WebSocket token-pipelining architecture streaming LLM tokens directly to TTS, cutting latency to 1.3–1.9s avg. Built stateful session management with VAD turn-taking & RMS echo suppression.
Multi-object tracking pipeline (YOLOv11 + ByteTrack) with cross-camera identity persistence & Vision LLMs for behavioral threat detection. Thread-safe FastAPI backend with 24-hour auto-purge privacy enforcement.
Enterprise-grade mock interview platform streaming STT → LLM → TTS via bidirectional Socket.IO using Groq Llama-3.3-70B (~200ms inference). Offloaded evaluation to async BullMQ workers with Redis caching.
Conversational AI agent featuring a 5-node LangGraph pipeline (router → reasoner → tools → editor) with ChromaDB domain lookup and a 25-case golden-set eval harness achieving a 96% pass rate.
Combining 5-node LangGraph logic, WebSocket STT-LLM-TTS streaming, and ChromaDB vector search.
Consistent algorithmic problem solver with strong proficiency in Data Structures, Graph Algorithms, Dynamic Programming, and High-Performance Optimizations.
Memoization, Tabulation, Knapsack, Bitmask DP, and Sequence Alignment.
BFS/DFS, Dijkstra, Topological Sort, Union-Find (Disjoint Set), and LCA.
Sliding Window, Binary Search on Answer Space, Monotonic Queues/Stacks.
B.Tech in Computer Science & Engineering — GPA: 8.3 / 10
Coursework: Data Structures & Algorithms, Operating Systems, Computer Networks, DBMS, System Design, OOP.
B.Sc in Data Science & Applications — GPA: 7.25 / 10
Coursework: Machine Learning, Statistical Inference, Python Data Science Stack, App Development.
97.81 percentile — Top 2% nationally among 1.1M+ candidates
Top 10 of 1000+ teams (Top 1%), BDH Implementation Problem
Finalist — Blockchain-based complaint management system from 1000+ applicants
3rd Runner-Up in national-level competitive development
330+ problems solved, Max Rating ~1600
FOSSology contributor (SQL logic & unit test suites) & GSoC 2026 applicant
