Final-year Electronics & Telecommunication engineer at VIT Pune who builds production GenAI systems end to end — RAG pipelines, LLM-powered products, and the full-stack surfaces that ship them. I do not stop at the model: Athena.ai and FairWork are deployed, wallet-authenticated, latency-measured systems, not notebooks.
My background is ECE, so almost every line of software I write is self-taught — which is why the range runs from LangChain and FastAPI to Solidity escrow contracts, SOC alert triage, and Siemens PLC machine vision on a live production line.
ROLE : AI/ML Engineer · Full-Stack Developer · Data Analyst
EXP : 4 internships · ~2 yrs hands-on (Jul 2024 → Mar 2026)
DOMAIN : GenAI/RAG · Cybersecurity (SOC) · Fintech Analytics · Industrial Automation
STACK : Python · FastAPI · LangChain · Next.js · React · PostgreSQL · Docker · AWS
OPEN_TO : AI/ML Engineering · Full-Stack Development · Data Analyticslanguages
ai · ml · genai
backend
frontend
databases
cloud · devops · web3
| Domain | Proficiency | Details |
|---|---|---|
genai/rag |
▰▰▰▰▰ | LangChain orchestration, pgvector retrieval on Supabase, Gemini + Whisper integration — 90% domain-routing accuracy on Athena.ai |
backend/apis |
▰▰▰▰▱ | FastAPI, Django REST Framework, WebSockets; refactored production endpoints for 20% lower response latency at Enovate IT |
full-stack/web |
▰▰▰▰▱ | Next.js 14 + React 18 + Tailwind, wallet auth via wagmi/RainbowKit, WebRTC, reusable component systems |
data/analytics |
▰▰▰▰▱ | Pandas, Statsmodels, SQL, Streamlit + Plotly, PowerBI — live Binance tick ingestion to OHLCV and z-score analytics |
security/soc |
▰▰▰▱▱ | SIEM triage of 80+ alerts/week, CTI log review across 500+ entries, Python automation cutting manual effort 35% |
web3/contracts |
▰▰▰▱▱ | Solidity + Foundry escrow on Polygon Amoy, XMTP messaging, full contract-flow validation via Postman |
▸ FairWork — decentralised freelance marketplace with on-chain escrow
Wallet-authenticated freelance platform where payment is held in a Solidity escrow contract until delivery is confirmed, with in-app messaging, video calls, and an AI-assisted dispute-analysis pipeline that produces structured reports for contested jobs.
| Stack | Next.js 14 · React 18 · Tailwind · wagmi + RainbowKit · FastAPI · Supabase · Solidity (Foundry) · Polygon Amoy · XMTP · Jitsi/JaaS |
| Scale | Full escrow lifecycle on testnet — fund, deliver, release, dispute |
| Impact | AI dispute pipeline generates structured verdict reports; REST + contract flows validated across edge cases in Postman |
| Links | Repository · Live Demo |
▸ Athena.ai — multi-domain RAG assistant (finance · health · education)
Full-stack Retrieval-Augmented Generation assistant that classifies an incoming query by domain, routes it through domain-specific retrieval, and answers from a pgvector-backed knowledge store — with voice-to-text input on the frontend.
| Stack | Python · LangChain · FastAPI · Supabase (pgvector) · Gemini API · React · Next.js |
| Scale | 3 knowledge domains, structured NLP routing layer |
| Impact | 90% domain-classification accuracy at sub-300ms API latency |
| Links | Repository · Live Demo |
▸ Quant Analytics Dashboard — real-time market data pipeline
Ingests Binance Futures tick data over WebSockets, normalises and persists it to indexed SQLite, resamples into multi-timeframe OHLCV, and computes pair-trading statistics behind an interactive dashboard.
| Stack | Python · Pandas · Statsmodels · SQL/SQLite · WebSockets · Streamlit · Plotly |
| Scale | Continuous live tick ingestion across multiple timeframes |
| Impact | Spread, hedge ratio, z-score, rolling correlation, volatility and stationarity analytics with threshold-based alerting |
| Links | Repository |
▸ CommunityAI — meeting transcription & content moderation platform
Whisper-powered speech-to-text service that turns raw meeting audio into minutes and action items, paired with a tuned multimodal moderation layer surfaced through a React dashboard and a Discord bot.
| Stack | Python · FastAPI · OpenAI Whisper · NLP · React · Discord API |
| Scale | Text, image, audio and video moderation paths |
| Impact | 90% accuracy on minutes/action-item extraction · 70% reduction in manual moderation workload |
| Links | Repository |
▸ Earlier builds
| Project | Year | Stack | What it does |
|---|---|---|---|
| Financial & Insurance Humanoid Agent | 2024 | Next.js · Python · TensorFlow · Flask · WebRTC | Multi-channel AI financial assistant with real-time market analysis and NLP query handling |
| Oil Spill Detection | 2024 | Python · TensorFlow · AIS data · OpenStreetMap API · Flask | Marine monitoring — vessel tracking, anomaly detection and environmental impact assessment |
| Road Repair Prioritization System | 2023 | Java · WEKA · JavaFX | ML-driven maintenance scheduling with an urban-planning decision interface |
- Contributed to a live production codebase building backend APIs with Python, Django REST Framework and FastAPI.
- Extended 5+ existing features against real product requirements.
- Debugged and refactored 3 API endpoints, cutting average response time by 20%.
- Operated inside agile sprint cycles — daily standups, sprint planning, on-schedule delivery.
- Monitored and triaged 80+ security alerts per week inside a live Security Operations Centre using SIEM tooling.
- Covered multiple threat categories, escalating by severity and category.
- Applied Cyber Threat Intelligence principles across 500+ log entries reviewed during the internship.
- Wrote Python automation scripts for log review, reducing manual effort by 35%.
Oct 2024 — Nov 2024 · Business Development Outreach Consultant Intern, Grant Thornton China — Remote
- Ran strategic market research and client opportunity mapping using the Power Interest Grid framework.
- Built a PowerBI client-analysis dashboard for real-time business metric visualisation.
- Applied Python and SQL analytics to develop data-driven retention strategies.
- Delivered stakeholder presentations incorporating machine learning insights for market segmentation.
- Implemented an automated gearbox assembly inspection system using Cognex Vision for part-presence detection.
- Integrated the vision system with a Siemens S7-1200 PLC for OK/NO-OK discrimination on a production line.
- Designed a real-time control panel interface with live data visualisation for operator monitoring.
- Engineered a vision-PLC interlocking safety mechanism.
| Achievement | Issuer / Venue | Year |
|---|---|---|
| Lead Author — "An Experimental Study on Solar-Based Green Hydrogen Production" | AIP (American Institute of Physics) | 2024 |
| Deep Learning Certification | NVIDIA DLI | 2024 |
| Advanced Quantum Computing (Qubit by Qubit) | The Coding School / Google-sponsored | 2024 |
| DevOps and Software Engineering | Coursera / IBM | 2025 |
| McKinsey Forward Program | McKinsey & Company | 2026 |
| Data Visualization | LinkedIn Learning | 2026 |
| Python for Data Professionals in Finance | LinkedIn Learning | 2026 |
Coursework: OOP · Data Structures & Algorithms · Advanced DSA · Databases · Discrete Mathematics · Operating Systems · Computer Networks · Machine Learning · Computer Vision · Embedded Systems
building:
- Athena.ai # multi-domain RAG assistant — finance, health, education
- FairWork # on-chain escrow marketplace with AI dispute analysis
- portfolio # siddhantswan.vercel.app
learning:
- production LLM systems # retrieval quality, latency, evaluation
- quantum computing # Qiskit and Cirq
exploring:
- vector retrieval at low latency # pgvector, hybrid search
- smart-contract architecture # Solidity, Foundry, Polygon
- real-time data pipelines # WebSockets to analytics
open_to:
- AI/ML Engineering
- Full-Stack Development
- Data Analytics
// ship the whole system, not just the model.