Publication-grade engineering specification for the full portfolio platform: Next.js shell, AI Financial Analyst, MLOps APIs, computer vision pipelines, NLP services, persistence, reliability controls, and print/export behavior.
Author: Anish Reddy
Role: AI Researcher and Systems Engineer
Web: https://anishreddy.online
GitHub: https://github.com/anishreddy13
Generated: 2026
0. Audit Scope and Ground Truth
Code-level review
This is a technical specification, not a resume. The document is derived from the repository files listed below and describes real system boundaries, API contracts, model paths, fallback behavior, deployment targets, and user-facing workflows. Where frontend marketing copy and live backend capability differ, this blueprint calls out the distinction instead of flattening it.
src/components/Navbar.tsx
src/components/Hero.tsx
src/components/About.tsx
src/components/Projects.tsx
src/components/FinancialAnalystExplainer.tsx
src/components/HeadlessFinancialAnalyst.tsx
src/components/PlantDiseaseDetector.tsx
src/components/InterviewAnalyzer.tsx
src/components/SocialSentinel.tsx
src/components/Contact.tsx
src/data/projects.ts
src/lib/financialAnalystClient.ts
src/lib/financialAnalystContent.ts
app/api/contact/route.ts
ml-backend/main.py
ml-backend/plant_inference.py
middleware.ts
1. End-to-End Infrastructure and Dataflow
System map
+-------------------------+ +------------------------------------+
| User Browser | | Next.js App Router / React Client |
| desktop / mobile | ----> | Navbar, Hero, Projects, ML Lab, |
| print preview | | AI Analyst, Contact, Print Export |
+-------------------------+ +------------------------------------+
|
v
+------------------------------------------------------------------------+
| middleware.ts / next-intl routing |
| matcher excludes /api, _next, _vercel, and static assets from i18n flow |
+------------------------------------------------------------------------+
| | |
v v v
+------------------+ +-----------------------+ +----------------------+
| Vercel API | | Hugging Face Space | | Render FastAPI |
| /api/contact | | Gradio AI Analyst | | ml-backend/main.py |
| validation + DB | | @gradio/client calls | | typed ML endpoints |
+------------------+ +-----------------------+ +----------------------+
| | |
v v v
+------------------+ +-----------------------+ +----------------------+
| Supabase table | | Groq / yfinance / | | model pickle files, |
| contact_submits | | pandas / PyTorch | | plant checkpoint |
+------------------+ +-----------------------+ +----------------------+
|
v
+------------------+
| Resend email |
| optional dispatch |
+------------------+
The analyst UI in HeadlessFinancialAnalyst.tsx coordinates a ticker-centered workflow. On Run Analysis it starts independent non-LLM service calls for forecast, financial tables, social sentiment, and neural chart vision, then performs sequential analysis and explainability calls. The shared client in financialAnalystClient.ts connects to the Hugging Face Gradio Space by ID, falls back to direct URL connection, retries failed calls, and normalizes cold-start errors.
The live/explainer architecture documents a five-role analyst pipeline: Researcher, Quant, Auditor, Competitor, and Editor. Older frontend metadata still lists a smaller three-agent story, so the blueprint treats five roles as the current flagship capability and the legacy content data as presentation metadata that should stay synchronized.
Neural Chart Vision
The neural vision feature is exposed through /get_neural_chart_vision. The explainer describes OHLCV windows, RSI/MACD/Bollinger/volume indicators, a candlestick image branch, a Transformer time-series branch, fusion logits, pattern labels, future-return estimates, and a deterministic fallback when trained weights are not deployed.
Portfolio Rebalancer
The implementation is an AI risk and exposure analysis engine, not a guaranteed quadratic optimizer. It evaluates allocation payloads, sector concentration, beta exposure, and then uses Llama reasoning to produce a reallocation strategy with graceful fallback behavior.
Service Reliability
/get_service_health checks runtime module readiness for market data, watchlist, rebalancing, neural chart vision, and the deep learning bundle. The frontend health strip summarizes readiness without triggering expensive external analysis requests.
3. Computer Vision and Deep Learning Pathology Suite
plant_inference.py loads an EfficientNet-B0 checkpoint from ml-backend/models/plant_disease/best_model.pt and class mapping JSON. The project card reports 97.84% accuracy, 20K+ dataset scale, and currently displays 15 classes.
Skin/Cancer Surface
main.py exposes /predict/cancer for structured cancer features and the portfolio project describes a skin disease image demo with seven clinical-style classes and top-5/risk UI. Treat those as separate surfaces unless the backend model path is unified.
Explainability
The PlantDiseaseDetector UI has heatmap/original attention views. The backend plant route returns classification payloads; any Grad-CAM overlay route should be verified before promising exported overlays in external documentation.
4. NLP, Security, and Speech AI Suite
Applied ML
Social and News Sentinel
SocialSentinel.tsx filters All, news, social, and video items for the selected stock. It expects source, source_type, content, sentiment, URL, and verification flags from /get_social_sentiment.
Sentiment / Spam / Emotion APIs
FastAPI endpoints /predict, /predict/spam, and /predict/emotion call loaded ML pipelines when pickle files are available. Responses include sentiment labels, spam probability and keywords, or 28-emotion ranked outputs plus demographic estimates.
Interview Analyzer
InterviewAnalyzer.tsx is browser-native. It uses speech recognition/transcript state, filler-word detection, WPM, vocabulary richness, sentence metrics, and rule-based coaching feedback without depending on a remote LLM call.
Project Catalog Mapping
src/data/projects.ts indexes Plant Disease, AI Financial Analyst, AI Interview Analyzer, Skin Disease Detection, Emotion Detection, Spam Detection, and Sentiment Analysis with route targets, status, technology tags, and feature metrics.
5. Backend Microservices, Schemas, and Runtime Controls
MLOps
Surface
Contract
Implementation
Reliability Notes
Contact API
POST /api/contact: name, email, subject, message, company honeypot
app/api/contact/route.ts validates payload, rate-limits by IP, stores to contact_submissions, and dispatches Resend email when configured.
src/lib/financialAnalystClient.ts connects by Space ID first and URL second, then retries failed predictions.
Client resets stale connections and maps app-config/503 errors to user-safe service messages.
Supabase Contact Schema
CREATE TABLE public.contact_submissions (
id UUID DEFAULT gen_random_uuid() PRIMARY KEY,
name TEXT NOT NULL,
email TEXT NOT NULL,
subject TEXT,
message TEXT NOT NULL,
created_at TIMESTAMPTZ DEFAULT clock_timestamp() NOT NULL
);
FastAPI Startup Strategy
main.py uses an async lifespan hook, optional loader functions, CORS middleware, health checks, and model-path validation. Missing models are recorded in OPTIONAL_IMPORT_ERRORS and endpoint requests return controlled 503 errors instead of terminating the service.
6. Print Export Architecture
PDF safe
The printable blueprint is mounted once in PlatformShell.tsx and hidden during normal browsing. Print actions from Navbar and Footer call window.print(). The @media print stylesheet isolates #print-root, hides every other element, forces white backgrounds and black text, disables fixed decorative overlays, and uses print-card containers with break-inside avoidance. Sections avoid min-h-screen and hard page breaks so the document flows densely without blank black or white pages.
Available for work · 2026
ML&AI
ENGINEER
Building intelligent systems where machine learning meets real products — from model training to production deployment.
50+
Students Mentored
mentorship · ml
2+
Yrs Building ML
3mo industry · frontend
4y
Teaching B.Tech
academia · labs
Open to work
ML & AI
50+
Students Mentored
mentorship · ml
2+
Yrs Dev & ML
3mo industry · frontend
4y
Teaching B.Tech Era
academia · labs
Scroll
SYSTEM TELEMETRY
AI FINANCIAL ANALYST:5 AGENTS ACTIVE
NEURAL CHART VISION:CNN + TRANSFORMER READY
PLANT DISEASE MODEL:EFFICIENTNET-B0 ONLINE
ML PRODUCTION BACKEND:FASTAPI / 45ms LATENCY
PIPELINE LOGS:SUPABASE REALTIME SYNCED
Featured AI ProductLive Headless API
MULTI-AGENT FINANCIAL AI
Enterprise-grade digital workforce. An orchestrated pipeline of 3 specialized AI agents executing research, quantitative analysis, and editorial synthesis to generate Bloomberg-style investment briefs.
LangGraphGroq (Llama-3)yfinanceDuckDuckGo SearchGradioHugging Face Spaces
I'm an ML & AI developer who builds end-to-end intelligent systems — from training models to shipping them into real products. I spent 4 years tutoring 50+ students during my B.Tech, which taught me how to break down complex problems clearly. Now I apply that same thinking to machine learning, backend systems, and whatever I'm obsessed with building next.
50+
Students\nMentored
2+
Yrs Building\nML
4y
Teaching\nB.Tech
3mo
Industry\nFrontend
inference · liveactive
PyTorch
LangChain
FastAPI
Next.js
02 / Technical Arsenal
The tools and frameworks I use to bring ideas to life.
Frontend
React
Next.js
TypeScript
Framer Motion
Three.js
Backend
Node.js
Python
FastAPI
PostgreSQL
Redis
Cloud & DevOps
AWS
Docker
Kubernetes
CI/CD
Terraform
ML & AI
PyTorch
TensorFlow
LangChain
OpenAI
HuggingFace
EXPERIENCE
Senior Full-Stack Engineer
Nexus Labs
2024Current
Lead Frontend Developer
Orbital Studio
2022
Full-Stack Developer
Pixel Foundry
2020
Junior Developer
CodeVault
2019
Academic Journey
Formal education laying the foundation for advanced machine learning, artificial intelligence, and software engineering.
Teegala Krishna Reddy Engineering College (TKREC)
Jan 2021 – Jun 2025
Bachelor of Technology - BTech, Artificial Intelligence and Machine Learning
