Adith Jose

Frontend • Senior • Vellore, India • UTC+05:30 • Work permits: India • Available now

Artificial Intelligence Specialist

Python

TypeScript

Node JS

UX & Visual Polish: 5/10

Responsive & Cross-browser: 4/10

Active 3 days ago

Invite to interview

Download CVCV

Overview

Technical skills

Timeline

Roles

Overview

Fullstack web developer (senior-level judgment in backend and integration, ~5+ years equivalent) focused on building recruitment and social-graph features with practical service layering. The strongest proven skill is backend service design and graph/recommendation logic as implemented in my-app/backend/src/services/friend-recommendation.service.ts. The codebase shows well-styled templates and scripts but lacks evidence of advanced frontend component architecture, automated a11y testing, or measured performance audits.

Phone

Technical skills

Languages

Python

TypeScript

Node JS

Express

Axios

Morgan

Helmet

Winston

Python

Pydantic

FastAPI

SQLAlchemy

Frontend

React.js

Tailwind CSS

Zod

React Router

Vite

Prettier

ESLint

AI/ML

LLM

huggingface_hub

OpenAI

ChatGPT

LangChain

Scikit-learn

Streamlit

Other

GitHub

autoprefixer

Rest API

Machine Learning

RAG

Embeddings

Google ADK

NLP

Recommender Systems

Structured Outputs

Timeline

Software Engineering Intern • Junior

Deutsche Telekom Digital Labs • Full-Time

Gurgaon In office

Developed capabilities for automated issue detection, service diagnostics, and intelligent workflow execution using Google ADK. Built search and embedding pipelines to support code retrieval backed by vector-based approaches. Focused on improving developer productivity through more relevant and faster information retrieval.

Google ADK

Embeddings

Senior Frontend Developer Confidence: Medium Fullstack

UI Component Architecture

3/10

How interface parts are built

Component structure is basic: a small React context and API surface exist, but most UI is template-driven HTML/CSS with limited reusable component boundaries or a custom component library.

Evidence

AI-RESUME-BOT-CAPSTONE/frontend/src/context/InterviewContext.ts

AI-RESUME-BOT-CAPSTONE/frontend/src/services/api.ts

smart-resume-screener/src/templates/dashboard.html

Responsive & Cross-browser

4/10

Works on all screens and browsers

Responsive practices are applied in plain HTML/CSS and Tailwind config; templates include media queries and fluid grids, and Tailwind setup enables dark mode and content scanning.

Evidence

smart-resume-screener/src/templates/dashboard.html

AI-RESUME-BOT-CAPSTONE/frontend/tailwind.config.js

AI-RESUME-BOT-CAPSTONE/frontend/src/index.css

Performance Optimization

3/10

Speed of the interface

Some backend-side optimization and caching strategies are present and BFS is implemented in-memory to reduce DB hits, but there is no measured RUM/bundle analysis or advanced client-side perf work.

Evidence

my-app/backend/src/services/friend-recommendation.service.ts: getCachedRecommendations / cacheRecommendations using RPC and DB caching

my-app/backend/src/services/friend-recommendation.service.ts: performBFS builds adjacency map in-memory to avoid repeated DB traversal

Accessibility & Semantics

2/10

Usable for everyone

Basic semantic HTML and labeled forms are present but there is little explicit a11y work (ARIA, keyboard management, automated a11y checks) in the human-authored UI artifacts.

Evidence

smart-resume-screener/src/templates/login.html: labeled form fields and focus styles

smart-resume-screener/src/templates/dashboard.html: form labels and structured content

State Management & Data Flow

3/10

Managing data in the app

Client-server data flow is straightforward (axios/api wrapper, React context) and backend shows service layering and error classes, but advanced client-side state discipline (cancellation, optimistic updates, state machines) is not evidenced.

Evidence

AI-RESUME-BOT-CAPSTONE/frontend/src/services/api.ts

AI-RESUME-BOT-CAPSTONE/frontend/src/context/InterviewContext.ts

my-app/backend/src/services/auth.service.ts

UX & Visual Polish

5/10

Look and feel quality

UX-focused polish is visible in CSS and templates (skeleton/empty states, animations, clear loading messages), producing a well-styled, usable interface even if it's template-driven rather than componentized.

Evidence

AI-RESUME-BOT-CAPSTONE/frontend/src/index.css: animations, card, button, video container styles

smart-resume-screener/src/templates/dashboard.html: empty states, loading placeholders, user-facing alerts

Expertise

React• Middle

HTML & CSS• Middle

Frontend Architecture & Build Tools• Middle

Industries

Professional Services• Middle

Technologies

TypeScript• Senior

Tailwind CSS

React.js

Vite

ESLint

Prettier

Zod

React Router

autoprefixer

Recommendations

Lead development of a recruitment-focused web product that needs robust server-side logic and integrative frontend work, owning the API, recommendation algorithms and the SPA client.
Implement a React component library and convert the template pages into composable, accessible components so the UI architecture scales and a11y can be audited programmatically.
Build a BFF or API gateway layer to centralize caching, request shaping, and add request cancellation/optimistic update patterns for a smoother client experience.
Harden seeding and auth flows by removing sample credentials from scripts and adding integration tests and CI checks for security and data handling.

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

my-app AI-RESUME-BOT-CAPSTONE smart-resume-screener

Middle AI/ML Engineer Confidence: Low LLM Engineer

Docu-focused LLM engineer (middle level) who builds practical local RAG systems and REST APIs. The strongest proven skill is building document ingestion and retrieval pipelines with page-level PDF chunking and metadata as shown in indexer.py and retriever.py. There is little evidence of custom model training, experiment tracking, production MLOps, or computational optimizations in public code.

Model Architecture & Training

1/10

How well models are designed and trained

Minimal model architecture or training work; mainly uses hosted/third-party models (Ollama/Groq) for embeddings and chat without custom architectures or training loops.

Evidence

DocuMind/indexer.py: ollama.embeddings calls for chunk embeddings

DocuMind/retriever.py: ollama.chat used to generate structured output

Data Pipeline & Feature Engineering

4/10

How data is prepared for models

Clear, practical document ingestion and chunking pipeline with PDF page-level metadata and chunk overlap, suitable for retrieval tasks.

Evidence

DocuMind/indexer.py: simple_chunk_text and index_document handling PDFs and per-page metadatas

DocuMind/test_workflow.py: end-to-end upload and chunk preview demonstrating pipeline behavior

Experimentation & Evaluation

2/10

How results are measured and tested

Basic integration-level testing and manual checks present, but no formal experiment tracking, validation metrics, or reproducible evaluation pipelines.

Evidence

DocuMind/test_workflow.py: script exercises upload and query endpoints as an integration demo

RAG/notebook/document.ipynb: EmbeddingManager with prints and basic checks (not a tracked experiment)

MLOps & Deployment

3/10

How models are shipped to production

Simple deployment-ready elements: a FastAPI app, persistent Chroma client, and structured Pydantic responses, but lacking production-grade MLOps features like monitoring, CI, containerization or versioned model management.

Evidence

DocuMind/api.py: FastAPI endpoints /index and /ask using Pydantic models for structured responses

DocuMind/retriever.py and DocuMind/indexer.py: chromadb.PersistentClient usage for persistent vector store

Computational Efficiency

1/10

How efficiently computing resources are used

No demonstrated computational efficiency work; embeddings are generated sequentially per chunk with no batching, quantization, or profiling evidence.

Evidence

DocuMind/indexer.py: for loop calling ollama.embeddings per chunk without batching

Research Depth & Innovation

1/10

Depth of research and new ideas

No evidence of original research, custom architectures, novel algorithms, or reproduced paper implementations; work is engineering-focused using existing models and frameworks.

Evidence

DocuMind/retriever.py and Langchain structured output examples (structured_op_pydantic.py) show applied use of models rather than research artifacts

Verified artifacts

AJ58282/fast-api-ecom• PR

Expertise

RAG• Middle

Document Intelligence & OCR• Middle

Industries

Artificial Intelligence• Middle

Technologies

LangChain

ChatGPT

Embeddings

Scikit-learn

NLP

huggingface_hub

LLM

RAG

Google ADK

Streamlit

OpenAI

GitHub

Structured Outputs

Recommender Systems

Machine Learning

Recommendations

Harden the ingestion pipeline: validate and sanitize uploaded filenames, handle large files safely, and add robust temp-dir cleanup and error handling in api.py.
Improve production readiness: add CI, unit/integration tests for indexer/retriever, Dockerfile, and a simple deployment manifest (e.g., docker-compose or Kubernetes) for reproducible runs.
Add evaluation and observability: implement retrieval/evaluation metrics (precision@k, recall), logging/metrics for latency and errors, and simple monitoring/alerts for the vector store and Ollama availability.
Optimize inference throughput: batch embeddings, add concurrency control or worker queue for indexing, and explore quantization or model selection to reduce latency and cost.

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

DocuMind Langchain RAG

Middle Backend Developer Confidence: Low API Engineer

Backend API engineer (mid-level) specializing in building AI-backed services and social-network backends with pragmatic, production-oriented code. The strongest proven skill is designing and implementing API-driven AI/ML features and RAG flows, evidenced by DocuMind (indexer.py, retriever.py, api.py) and the AI Resume Bot services (question_generator.py, analysis_service.py). There is limited evidence of formal migration history, load-testing, or mature observability/operations tooling in public code.

API Design

5/10

How well APIs are designed

APIs are well-structured with typed request/response models, consistent HTTP error usage, and some rate-limiting/auth middleware integration, but there is no visible global versioning strategy, idempotency key handling, or advanced pagination conventions.

Evidence

newuserAJ/DocuMind/api.py: FastAPI endpoints (/index, /ask) with Pydantic QueryRequest and response_model newuserAJ/AI-RESUME-BOT-CAPSTONE/backend/app/routes/interview.py: robust FastAPI route checks and HTTPException usage newuserAJ/my-app/backend/src/routes/graph.ts: Express route handlers with requireAuth and generalRateLimit middleware

Data Layer & Database

4/10

Working with databases

Data layer shows async SQLAlchemy models and deliberate Supabase usage with upserts and RPC calls; seeders and careful error handling are present, but there is no visible migration history or explicit transaction/isolation tuning.

Evidence

newuserAJ/AI-RESUME-BOT-CAPSTONE/backend/app/database.py: async SQLAlchemy engine and session setup newuserAJ/my-app/backend/scripts/seed-sample-data.ts: explicit seeding with Supabase admin and upsert usage newuserAJ/my-app/backend/src/services/friend-recommendation.service.ts: use of RPC get_cached_recommendations and upsert for caching

Scalability & Performance

3/10

Handling load and speed

Some scalability considerations exist (chunking for embeddings, caching of recommendations, rate-limit middleware), but there is limited evidence of invalidation strategies, measured performance tuning, load-testing artifacts, or sophisticated connection pooling configuration.

Evidence

newuserAJ/DocuMind/indexer.py: chunking strategy and batching embeddings before storing in ChromaDB newuserAJ/my-app/backend/src/services/friend-recommendation.service.ts: caching recommendations with expiry and cleanup RPC newuserAJ/my-app/backend/src/middleware/rateLimiting.ts: rate limit middleware referenced (package.json includes express-rate-limit)

System Architecture

4/10

Overall system structure

Clear modular service and router separation (services/, routes/, utils/), and separate AI, indexer, retriever components for RAG, indicating deliberate boundaries; architecture remains single-process services rather than multi-service distributed design.

Evidence

newuserAJ/DocuMind: separation between api.py, indexer.py, retriever.py, models.py newuserAJ/AI-RESUME-BOT-CAPSTONE/backend/app/services/: question_generator, analysis_service, resume_parser newuserAJ/my-app/backend/src/: services, routes, middleware folders (e.g., auth, rateLimiting, graph routes)

Security & Auth

4/10

Protecting data and access

Authentication and authorization flows are implemented with care (token refresh, password reset, synthetic-email phone flow), plus security dependencies present, but there are some insecure sample artifacts (plain passwords in seeders) and no evidence of dependency vulnerability audits or secrets rotation policies.

Evidence

newuserAJ/my-app/backend/src/services/auth.service.ts: signUp, logIn, refreshToken, forgotPassword implementations and syntheticEmailFromPhone handling newuserAJ/my-app/backend/package.json: security-related middleware and libs (helmet, express-rate-limit, jsonwebtoken, morgan, winston) newuserAJ/my-app/backend/scripts/seed-phone-basic.ts: sample passwords and synthetic email use (security hygiene risk)

Reliability & Observability

4/10

Stability and monitoring

Good pragmatic reliability: try/catch handling, logging, cleanup of temp files, and defensive DB error handling; observability is basic (logger usage) but lacks structured metrics, retry-with-backoff, circuit breakers, and documented graceful shutdown patterns.

Evidence

newuserAJ/DocuMind/retriever.py: structured exception handling for missing collection and JSON decode fallback newuserAJ/DocuMind/api.py: temp file creation and cleanup in index_file endpoint (finally block) newuserAJ/my-app/backend/src/services/friend-recommendation.service.ts: consistent logging and try/catch blocks for reliability

Expertise

Backend AI & LLM• Middle

Databases & Vector Storage• Middle

Node.js• Middle

Python• Middle

Industries

Education• Middle

Technologies

Python• since 2025 • Middle

Node JS• Senior

Rest API

Express

SQLAlchemy

FastAPI

Winston

Pydantic

Axios

Helmet

Morgan

Recommendations

Extend and harden the RAG pipeline for production: add batching controls, backpressure, cache invalidation, and integration tests around index/query flows.
Improve operational readiness: add structured metrics, retry-with-backoff for external calls, graceful shutdown hooks, and documented migration scripts for the databases.
Expand security hygiene: remove/replace hard-coded sample credentials, add secrets management, and include dependency vulnerability scanning in CI.
Productize the friend-recommendation system: add unit/integration tests for BFS and scoring logic, and benchmark/cost the Supabase queries for large graphs.

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

DocuMind my-app AI-RESUME-BOT-CAPSTONE