SOFTWARE DEVELOPER · AI/ML

Hi, I'm Het Patel

I build trading systems
& full‑stack platforms.

Final‑year ICT engineer with a minor in Computational Data Science - interned at Trionic Technologies and working on IATS.

Het Patel
</> PROFILE LIVE
const engineer = {
  name: "Het Patel",
  role: "Software Developer",
  "Leetcode rating": 1721,
  status: "open for work"
};
01 · LISTING PROSPECTUS

Education

Two share classes under one issuer.

Pandit Deendayal Energy University
Aug 2022 - May 2026
8.80
CGPA / 10.00
02 · OPEN BELL

Experience

Current position, reported as of this quarter.

Trionic Technologies LLP

Software Developer Intern - AI/ML
Jan 2026 - Present
Ahmedabad, India
4
AGENTS IN ARCHITECTURE
15+
RISK CONTROLS LIVE
67+
PYTEST CASES
3
EXECUTION MODES
  1. Designed and shipped an intraday agentic NSE trading system where LLMs only propose trades; schema validation, sanity checks, and a deterministic RiskGovernor approve / reject / halt before any order reaches Zerodha via Kite MCP, with DRY_RUN / PAPER / LIVE gates.
  2. Built the news → ticker → NewsRank pipeline (NSE/BSE + RSS + Yahoo): nil‑filing filters, symbol canonicalization (e.g. LIC→LICI, BSE equity vs gold‑bond noise), taxonomy scoring with recency decay, and single‑pick execution under session risk / MIS limits - backed by 15+ live risk controls.
  3. Added an offline safeguards backtest (RiskGovernor + Yahoo OHLC) reporting win rate and wallet return (e.g. RELIANCE ~37% win rate, ~+6% return), plus a hybrid YAML+Python harness with 67+ pytest cases across risk, API, and trading workflows.
LIVE TICKET TAPE · KITE ORDERS

Personal Zerodha Kite order / P&L screenshots from system‑driven MIS fills. Internal IATS UI omitted per company policy.

ATHERENERG MIS buy and sell complete orders on Zerodha Kite
ATHERENERG · ROUND TRIP - MIS LIMIT buy @ 1,456.40 (11:05) → sell @ 1,471.20 (12:02); both legs COMPLETE.
SAREGAMA trade with stop-loss fill and cancelled target order
SAREGAMA · SL + TARGET - entry @ 508.90; stop‑loss sell COMPLETE @ 506.65; profit target @ 514.15 CANCELLED after SL fill.
Zerodha Kite positions total P and L plus 8.65 for the session
SESSION P&L - Total P&L +8.65 with ATHERENERG +10.90 and SAREGAMA −2.25 closed MIS (qty 0) on a red NIFTY day.
03 · PORTFOLIO HOLDINGS

Projects

Two open positions. Click a holding to view full disclosure.

POS‑01 · ALPHA
VMD + BiLSTM Short‑Term Forecasting
DEEP LEARNING

Built an end‑to‑end 5‑minute equity forecasting pipeline on NSE data: ingest 1‑minute OHLCV, resample, engineer technicals, filter features by correlation, then forecast with LSTM / BiLSTM baselines and a VMD-BiLSTM hybrid. Close price is decomposed into 4 modes, each predicted with BiLSTM and summed back - the hybrid beat every baseline on the holdout set.

SmartAPIVMDBiLSTMLSTMTensorFlowpandas_taPython
  • Ingested multi‑year 1‑minute bars, resampled to 5‑minute OHLCV, and trained with a 24‑bar lookback, 70/15/15 split, EarlyStopping + best‑checkpoint.
  • Engineered indicators (SMA/EMA/WMA/KAMA, Bollinger, RSI, MACD, ADX, ATR, OBV, lags) and kept only high‑correlation predictors (|ρ| ≥ 0.85 with close).
  • Benchmarked univariate & multivariate LSTM / BiLSTM against VMD-BiLSTM (K=4 IMFs); the hybrid reconstructed price and outperformed all baselines on R² / MAE / RMSE / MAPE.
VIEW FULL DISCLOSURE
POS‑02 · CORE
Prototerra
FULL‑STACK PLATFORM

ProtoTerra is a scalable, 3‑tier e‑commerce architecture featuring a Server‑Side Rendered (SSR) Next.js storefront, a dedicated Next.js Admin portal, and a secure Node.js / MongoDB centralized API. The frontend leverages a custom HTML5 Canvas engine and Lenis scroll‑mapping for a highly optimized 60FPS cinematic experience. The backend keeps data integrity for up to 1,000 concurrent users with automated webhook pipelines, Mongoose atomic transactions to prevent inventory overselling, and HMAC‑SHA256 authenticated Razorpay checkout flows.

Next.jsNode.jsMongoDBCanvasLenisNextAuthRazorpay
1,000
CONCURRENT USERS · k6
94%
ASSET PAYLOAD CUT
60FPS
300vh CANVAS SCROLL
  • Engineered a scalable full‑stack e‑commerce architecture (Next.js, Node.js, MongoDB) and optimized data fetching with compound indexing, successfully supporting sustained loads of 1,000 concurrent users validated via k6 stress testing.
  • Slashed animation asset payloads by 94% (from 140MB to <8MB) by engineering a custom HTML5 Canvas rendering engine, replacing 160 heavy DOM elements to achieve seamless 60 FPS across a 300vh interactive scroll.
  • Architected a frictionless auth flow with NextAuth.js for Google OAuth and secure JWT session management, bridging secure backend access with rapid e‑commerce checkout.
DEMO REEL · LOOPING
DESKTOP · SSR STOREFRONT - custom Canvas engine + Lenis scroll‑mapping across a 300vh cinematic hero; 60FPS product storytelling before checkout.
MOBILE · RESPONSIVE UI - same Canvas / scroll system on a phone viewport; touch‑friendly browse → cart → Razorpay‑ready flow.
VIEW FULL DISCLOSURE
04 · RANKINGS

Achievements

Competitive programming standing.

Green = rating gain, red = rating drop, gold ring = peak. Hover a candle for detail.
Green = rating gain, red = rating drop, gold ring = peak. Hover a candle for detail.
05 · WATCHLIST

Technical Skills

Instruments held, grouped by class.

LANGUAGES
▲ +1.5%
C++▲ +1.2%
Python▲ +3.4%
JavaScript▲ +2.1%
MySQL▼ -0.3%
MongoDB▲ +1.8%
HTML5 / CSS3▲ +0.9%
FRAMEWORKS & LIBRARIES
▲ +1.2%
Next.js▲ +2.6%
React.js▲ +1.5%
JWT▼ -0.4%
Tailwind CSS▲ +1.1%
TOOLS & PLATFORMS
▲ +0.5%
Git▲ +0.7%
VS Code▲ +0.5%
Jupyter▼ -0.2%
Google Colab▲ +1.0%
CONCEPTS
▲ +2.9%
DSA▲ +2.9%
OOP▲ +1.6%
AI / ML Fundamentals▲ +4.2%
06 · CLOSING BELL

Get in Touch

Markets close, but this line stays open.

ORDER TYPE: MESSAGESTATUS: DRAFT

CODEFORCEShetpatel12

Currently based in Ahmedabad, India. Graduating May 2026. Open to full‑time Software Development roles.