I'm a Software Engineer in Bengaluru who builds AI-powered products end to end, from the model and the backend to the interface people use.
At MapmyIndia (Mappls) I'm building a multi-agent AI platform for DRDO (CAIR): a router agent that understands natural-language questions and hands them to 10 specialist agents running on self-hosted LLMs with MCP tools and RAG.
Before that I spent a year at Fidelity Investments shipping ML models and decision tools for product teams. Outside work I build full-stack apps with React, Node.js and Python, including real-time multiplayer games, a messaging app and a crash-safe npm package.
What I do
AI Agents & Automation
Multi-agent systems with tool calling, MCP servers and RAG that turn manual workflows into a single natural-language interface.
Full-Stack Development
React frontends and Node.js / Python backends with REST APIs, WebSockets and real-time features.
Applied Machine Learning
Classification and prediction models, from feature engineering to serving them behind an API.
Databases & System Design
PostgreSQL, MongoDB, Redis and SQLite; data persistence, caching and crash recovery.
Contract via Brick Solutions · AI Automation & Geospatial Analytics Platform for DRDO (CAIR)
Designed a multi-agent architecture: a router agent classifies the intent of natural-language queries and dispatches them to 10 specialist agents (warning-area lookup, distance and proximity computations, ENC chart data, live web scraping, licensed data-feed APIs).
Built agents as Python tool-calling functions (the LLM sets parameters, deterministic code executes) and as Model Context Protocol (MCP) client-server tools on self-hosted LLMs with RAG.
Read-only agent interfaces over raw real-time data; cleaning, validating and consolidating multi-source outputs into reports and dashboards.
Apprentice, Analytics & ML – Fidelity Investments
Feb 2025 — Feb 2026
Digital Analytics and Machine Learning
Built a Customer Effort Score KPI and logistic-regression predictor (12 features); CES replaced NPS as the primary app-health metric.
Built a transaction intent classifier across 88 transaction types.
Built Streamlit decision tools for page prioritization (9 features, scenario simulation) and advisor ranking (16 features).
Router agent classifies natural-language queries and dispatches them to 10 specialist agents (geospatial computations, ENC chart data, web scraping, data-feed APIs) on self-hosted LLMs with MCP and RAG.
Automated pipeline that checks 60+ company career boards (Workday, Oracle, Greenhouse, Lever, Ashby) every hour, filters new roles by title, description and location, scores them against my profile with an LLM, and pushes the best matches to Telegram.
Personal project · private repo
SortOS – Privacy-First AI Desktop Agent
Python · Local LLM · Speech-to-Text · JSON Tool Calling
Offline voice-controlled desktop agent: local speech-to-text and TTS, LLM intent parsing with structured JSON tool calling, and whitelist-based command routing that separates understanding from execution.
Personal project
Three-State Counter (npm)
Node.js · SQLite · WAL
Crash-safe Node.js library with a RAM → WAL → SQLite persistence strategy, zero-loss sync mode and state recovery from logs.
Custom CNN trained on 61,000+ leaf images, served through a Flask API for real-time diagnosis.
Innosights internship
Customer Effort Score (CES)
Python · Scikit-learn · Pandas
New KPI for transaction friction plus a logistic-regression model on 12 engineered features; CES replaced NPS as the primary app-health metric.
Fidelity · internal
Transaction Intent Classifier
Python · Scikit-learn
Multi-class model predicting user intent across 88 transaction types to support workflow optimisation.
Fidelity · internal
Page Prioritization Tool
Python · Streamlit
Streamlit tool ranking app pages on 9 features (friction, revenue, volume) with scenario simulation to decide which pages move to the native app.
Fidelity · internal
Advisor Ranking
Python · Streamlit
Expanded advisor ranking from AUM-only to 16 strategic features to optimise incentive allocation, with correlation analysis for product teams.
Fidelity · internal
Contact
Open to Software Engineering, AI/ML and Data roles in Bengaluru or remote. The fastest way to reach me is
yashwanthchikkihd@gmail.com, or send a message below.