Software Engineer & AI Automation Specialist
I build the AI and backend systems that make businesses faster.
Backend systems, AI pipelines, and the automation that makes your business faster. Ask me about it — I answer directly.
01 · about
I'm Rushad Khandekar, a software engineer based in Navi Mumbai, India. I work on the invisible stuff that makes products work — backend systems, AI pipelines, and automation.
At Capgemini I built production AI for enterprise-scale problems. In my own time I build ambitious systems for fun — including an autonomous trading bot and a personal finance AI. I'm as comfortable explaining things in plain business language as I am reading code.
Python · FastAPI · LangGraph · RAG · LLMs · GCP · Docker
02 · experience
Software Engineer
Sep 2024 – Jul 2026Capgemini
Built production AI systems inside a large enterprise — an AI troubleshooting assistant that saves support engineers hours per issue, and a real-time order-accuracy system.
- ▸Designed an AI assistant (iAssist) that finds the root cause of production issues from a single ID — hybrid keyword + semantic search over millions of logs.
- ▸Shipped a real-time system for McDonald's Japan validating AI camera detections against live point-of-sale data, with sub-millisecond processing.
03 · selected work
iAssist
enterpriseAn AI troubleshooting agent I built at Capgemini that turns a single error ID into a recommended fix — by searching logs, past incidents, and expert knowledge.
+
iAssist
enterpriseAn AI troubleshooting agent I built at Capgemini that turns a single error ID into a recommended fix — by searching logs, past incidents, and expert knowledge.
Three-repo system: log ingestion into OpenSearch, scheduled embedding + index sync, and a FastAPI service orchestrating LangGraph workflows. Hybrid search (keyword + semantic) plus RAG grounding assemble logs, incidents, and SME knowledge before the LLM answers. Continuous learning loop re-indexes new resolutions daily.
Order Accuracy System
enterpriseA real-time validation system for McDonald's Japan that cross-checks what AI cameras see against what the register says — flagging mismatches instantly.
+
Order Accuracy System
enterpriseA real-time validation system for McDonald's Japan that cross-checks what AI cameras see against what the register says — flagging mismatches instantly.
Primary backend engineer. Flatbuffers binary telemetry from Sony IMX500 Edge AI cameras for sub-millisecond zero-copy serialization; TCP/IP socket listener for HubHook POS event streams (STORE/BUMP/SERVE); state-sync logic reconciling async POS events to prevent false flags; crew-facing dashboard with green/red per-order feedback.
TITAN Trader
personalAn autonomous intraday trading bot for Indian markets — it reads the morning market, debates which trades are worth it, and executes them live with hard risk limits it can't override.
+
TITAN Trader
personalAn autonomous intraday trading bot for Indian markets — it reads the morning market, debates which trades are worth it, and executes them live with hard risk limits it can't override.
7-stage morning plan pipeline over a 40-line YAML market context packet; 7 hard Python risk gates the LLM cannot override. Real-time WebSocket streamer builds 1-minute candles; full execution lifecycle (entry, ATR trailing stop, auto-close). Thesis invalidation engine auto-exits on 5 intraday checks.
TITAN 2.0
personalA personal finance brain that watches my stocks, mutual funds, and goals — then holds an AI debate about them once a week and alerts me when something needs attention.
+
TITAN 2.0
personalA personal finance brain that watches my stocks, mutual funds, and goals — then holds an AI debate about them once a week and alerts me when something needs attention.
Privacy-first advisory system. 4-agent AI debate (Bear, Bull, Visionary, Moderator) over a 2,000-token fact sheet; all numbers pre-computed by Python math engines (XIRR via SciPy, HHI concentration). 6-threshold rule engine sends instant Telegram alerts with zero LLM cost. Pure Python, no LangGraph, no vector DB.
04 · what I can build for you
AI chatbots & assistants
A bot that actually knows your business — trained on your own documents, pricing, and processes.
Automated data pipelines
Connect scattered tools and data sources into one automated flow — no more copy-pasting between spreadsheets.
RAG & enterprise search
Let your team ask questions and get grounded, sourced answers from your documents, tickets, and logs.
Backend APIs & integrations
The engine room — APIs and integrations that make your apps and tools work together reliably at scale.
Custom automation & bots
Telegram alert bots, analysis bots, any repeatable rule-based process — automated.
AI architecture & consulting
Not just code — a plan. The right AI/backend design before you spend on the wrong stack.
05 · contact
Let's build something
One email is all it takes. Tell me what you're trying to do — I'll tell you honestly whether I can help and how.
rushadkh2020@gmail.com · Navi Mumbai, India