Yuvraj Jaiswal
AI Engineer
Management: Under a year (1-5 people)
Python
TypeScript
Node JS
SQL
JavaScript
Data Pipeline & Feature Engineering: 6/10
MLOps & Deployment: 5/10
Research Depth & Innovation: 4/10
Active 3 days ago
Invite to interview
Download CVCV
Overview
Technical skills
Projects
Timeline
Roles
Overview
LLM Engineer (senior-level) specializing in production-grade retrieval-augmented generation and agentic workflows for developer-facing content tools. The strongest proven skill is engineering resilient RAG pipelines and retrieval logic as evidenced by ingestion/indexer.py (Chroma client, LRU eviction) and rag/retriever.py (MMR + MultiQueryRetriever). Not evidenced publicly are custom model training, distributed GPU optimization, formal experiment tracking (W&B/MLflow), or exhaustive automated test suites.
Phone
Technical skills
Languages
5
Python
TypeScript
Node JS
SQL
JavaScript
Node JS
5
Express
Bcrypt
Multer
Axios
Dotenv
Python
3
FastAPI
Requests
Pydantic
Databases
3
PostgreSQL
Redis
FAISS
AI/ML
12
Claude
LangChain
LLM
OpenAI
Pandas
LangGraph
Streamlit
RAG
AI Agents
GPT-4
Embeddings
Structured Outputs
Frontend
8
React.js
Tailwind CSS
React Query
Zod
React Router
Redux Toolkit
React Hook Form
Vite
Other
11
Docker
AWS
Rest API
GitHub
Tavily
DLP
Groq
CI/CD
Prompt Engineering
Agentic Workflows
Tool Use
Projects
TechScribe AI -
- Built an end-to-end GenAI application using LangChain/LangGraph, typed state, agentic workflows, document loaders, chunking,
- embeddings, retrieval, MultiQueryRetriever, and Reflexion for document-grounded responses.
- • Built a RAG evaluation harness across 100+ test queries using RAGAS for faithfulness, answer relevancy, and context precision;
- implemented source citations linking generated responses to repository files and source locations
Python
LangChain
LangGraph
Embeddings
Agentic Workflows
RAG
Ragas
AI Job Search Agent - LangChain Tool-Calling System
- Built an asynchronous Python AI agent using LangChain and Groq for job discovery, resume-aware filtering, tool/function calling, and
- Pydantic-validated structured outputs across 10+ attributes.
- • Designed an agentic workflow processing 50+ live job postings per search, separating retrieval, tool execution, filtering, and
- structured generation for reliable AI-assisted decision workflows
LangChain
Groq
Python
Pydantic
Tavily
FastAPI
AI Shopping Assistant — Budget-Constrained E-Commerce Agent
- Built a FastAPI + React AI assistant integrating LLM intent extraction, multi-store APIs, conversational state, external data, SQL-
- backed services, and budget constraints.
- • Separated LLM extraction from deterministic Python budget allocation and product ranking to improve consistency, debuggability,
- and token efficiency.
LLM
Python
FastAPI
React.js
Rest API
SQL
Timeline
Cloud Engineer
•
Middle
Erasmith Technologies Pvt. Ltd.
•
Full-Time
Built and maintained Python/FastAPI backend services and REST-based enterprise applications with PostgreSQL and Redis. Integrated OpenAI and Claude APIs and implemented RAG pipelines using LangChain with FAISS-style retrieval and structured LLM outputs. Developed KPI analytics and reporting microservices, implemented JWT/RBAC and structured logging, and deployed services in Docker with asynchronous processing and monitoring.
Python
FastAPI
LangChain
LangGraph
OpenAI
Claude
FAISS
RAG
Structured Outputs
PostgreSQL
Redis
Docker
Galgotia's College of Engineering and Technology
Bachelor's Degree •
Computer Science And Engineering
Senior AI/ML Engineer
Confidence: High LLM Engineer
LLM Engineer (senior-level) specializing in production-grade retrieval-augmented generation and agentic workflows for developer-facing content tools. The strongest proven skill is engineering resilient RAG pipelines and retrieval logic as evidenced by ingestion/indexer.py (Chroma client, LRU eviction) and rag/retriever.py (MMR + MultiQueryRetriever). Not evidenced publicly are custom model training, distributed GPU optimization, formal experiment tracking (W&B/MLflow), or exhaustive automated test suites.
Model Architecture & Training
2/10
How well models are designed and trained
No custom model training or novel architectures; the project uses hosted LLMs (ChatGroq) and prompt-based generation rather than training or modifying models.
Evidence
composer/generator.py: uses ChatGroq to generate first drafts
rag/retriever.py: uses ChatGroq for deterministic query expansion
Data Pipeline & Feature Engineering
6/10
How data is prepared for models
Robust ingestion and preprocessing pipeline with file filtering, chunking, cleaning, embedding, and vectorstore management; clear handling of large-repo heuristics and LRU eviction.
Evidence
ingestion/github_fetcher.py: fetch_repo_files, _should_include, _fetch_file_content with retry and file filters
ingestion/chunker.py: chunk_files (chunking logic present)
ingestion/indexer.py: build_vectorstore, _get_chroma_client, evict_lru_if_needed (Chroma persistence and LRU)
Experimentation & Evaluation
3/10
How results are measured and tested
Has an evaluation stage (RAGAS metrics) and UI display for metrics, but lacks experiment tracking or reproducible experiment management.
Evidence
evaluation/ragas_eval.py: evaluator integration and metric definitions
app/main.py: calls evaluate_output and renders groundedness/metrics
MLOps & Deployment
5/10
How models are shipped to production
Good practical MLOps/serving considerations for a retrieval pipeline: vectorstore lifecycle, cloud vs local clients, LRU eviction, re-ingest flows and a Streamlit UI; not a full production serving stack but solid operational engineering for RAG apps.
Evidence
ingestion/indexer.py: _ensure_cloud_database, _get_chroma_client, collection lifecycle and eviction
app/components/sidebar.py and app/main.py: re-ingest, load_vectorstore and session management for serving
Computational Efficiency
3/10
How efficiently computing resources are used
Some efficiency-aware choices (MMR, fetch_k tuning, request throttling, size limits) but no GPU/quantization/profiling or low-level optimization.
Evidence
rag/retriever.py: MMR settings, fetch_k multiplier and audience-based k
ingestion/github_fetcher.py: REQUEST_DELAY and MAX_FILE_BYTES to control API/rate and size
Research Depth & Innovation
4/10
Depth of research and new ideas
Shows thoughtful retrieval strategy, multi-query expansion and reflexion loop design using existing research concepts, but mostly implements existing methods rather than introducing novel algorithms.
Evidence
rag/retriever.py: multi-query retrieval + MMR re-ranking strategy
reflexion/graph.py and reflexion/chains.py: reflexion critique loop integration (LangGraph)
Expertise
RAG• Senior
Industries
Cybersecurity• Senior
Technologies
Python• since 2025 • Senior
SQL
PostgreSQL• since 2025
Redis• since 2025
LangGraph• since 2025
LangChain• since 2025
Claude• since 2025
FAISS• since 2025
Groq
Embeddings
Prompt Engineering
AI Agents
CI/CD
Pandas
AWS
Docker• since 2025
LLM
RAG• since 2025
Streamlit
Pydantic
Requests
OpenAI• since 2025
GitHub• since 2026
GPT-4
Structured Outputs• since 2025
Tavily
Agentic Workflows
Tool Use
GitHub• mentioned only
Recommendations
- Build production RAG/agent applications that need robust repo ingestion, citation-aware generation, and retrieval tuning.
- Implement backend services that integrate vectorstores and LLMs with operational concerns (LRU eviction, re-ingest, streamlit or lightweight UIs).
- Develop API clients and ETL pipelines for enterprise integrations, especially security/monitoring-focused tools like DLP export and normalization.
- Contribute to retrieval tuning and evaluation pipelines (expand RAGAS metrics, add reproducible experiment tracking and CI).
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Senior Backend Developer
Confidence: High API Engineer
Backend engineer (Senior) focused on Node.js-based internal productivity systems with strong operational safety practices. The strongest proven skill is careful database and data-migration design, demonstrated by the idempotent, transactional migration and data-migration pipeline in backend/src/migrations/001_analytics_timesheet_refactor.js. There is little public evidence of AI/LLM production work, message-driven architectures or large-scale distributed systems design in the code.
API Design
5/10
How well APIs are designed
Reasonable API design patterns and defensive behavior are present (clear controller/service separation, meaningful error returns and dedupe/upsert semantics), but there is little evidence of formal versioning, global idempotency keys or a documented API contract.
Evidence
Kanvance/backend/src/services/timesheet.service.js: exports.importRows, previewConflicts and commitToTimeLogs show request->validate->preview->commit flow
Kanvance/backend/src/controllers/project.controller.js: function canModify and isChangingOwner signatures imply guarded controller logic
Kanvance/backend/src/controllers/reports.controller.js: generateExcelReport shows a controller-level report generation endpoint
Data Layer & Database
7/10
Working with databases
Strong relational DB focus with a careful, idempotent migration chain, explicit transactions and FK/constraint awareness plus batch/deduplication logic to avoid duplicates.
Evidence
Kanvance/backend/src/migrations/001_analytics_timesheet_refactor.js: idempotent DDL guarded by INFORMATION_SCHEMA checks, transaction/rollback, dry-run and data-migration paths
Kanvance/backend/src/services/assignment.service.js: uses pool.getConnection with beginTransaction/commit/rollback for mutating operations
Kanvance/backend/src/services/timesheet.service.js: preloads projects and subtasks to avoid N+1 and uses parameterised pool.execute queries
Scalability & Performance
5/10
Handling load and speed
Some scalability and performance care (connection pooling, preloading, batch operations, file-locking for concurrent writers) but no evidence of horizontal scaling, caching strategy with invalidation, queueing or load-testing.
Evidence
Kanvance/backend/src/services/timesheet.service.js: pre-loads projects/subtasks and performs batched conflict detection to avoid N+1
YuvrajJais9257/Kanvance/backend/src/migrations/001_analytics_timesheet_refactor.js: DRY-RUN and bulk migrate modes to batch data migration
Kanvance/backend/src/services/timesheetExcel.service.js: acquireLock and atomicWrite implement concurrency control for file writes
System Architecture
6/10
Overall system structure
Clear modular structure (services, models, controllers, migrations, utils) and purposeful boundaries; manifests production operational concerns (dry-run flags, run logs, backups) though service decomposition remains monolithic rather than a multi-service topology.
Evidence
Kanvance/backend/: separated folders src/services, src/models, src/controllers, src/migrations and src/utils
Kanvance/backend/src/services/timesheetAutofill.service.js: orchestrator calling models/services and writing run logs (shows deliberate module boundaries)
Kanvance/backend/src/migrations/*.js: multiple migration scripts forming a migration_chain
Security & Auth
5/10
Protecting data and access
Good practical input-safety and integrity practices (sanitization, prepared statements, validation) and use of DB constraints, but little explicit evidence of a complete auth lifecycle, token refresh/revocation, secrets rotation or dependency vulnerability auditing.
Evidence
Kanvance/backend/src/services/timesheetExcel.service.js: sanitizeCell to strip control characters before writing
Kanvance/backend/src/services/timesheet.service.js: validateRows implements extensive row-level validation and permission checks
Kanvance/backend/src/migrations/001_analytics_timesheet_refactor.js: CHECK constraints and FK references in created tables
Reliability & Observability
7/10
Stability and monitoring
Strong reliability focus: transactional migrations with rollback, dry-run and data-migration isolation, atomic file write with backup/validation, file locking and run logs; fewer signs of automated retry policies, circuit breakers, or distributed tracing integration.
Evidence
Kanvance/backend/src/migrations/001_analytics_timesheet_refactor.js: explicit BEGIN/COMMIT, rollback on error, structured error logging and dry-run
Kanvance/backend/src/services/timesheetExcel.service.js: atomicWrite with backup, temp-file validation and acquireLock to handle concurrent writers
Kanvance/backend/src/services/timesheetAutofill.service.js: dry-run mode, appendRunLog and per-employee error handling that continues the run
Expertise
Node.js• Middle
Databases & Vector Storage• Middle
Industries
Professional Services• Middle
Technologies
Rest API
FastAPI• since 2025
Bcrypt
Multer
Dotenv
Recommendations
- Develop backend services for enterprise internal tooling and reporting pipelines where transactional safety and data migration are critical.
- Implement robust import/export and ETL features that require careful deduplication, idempotency and atomic writes (timesheet/excel-style pipelines).
- Build database migration and data-reconciliation tooling (dry-run, audit logs, error logging) for MySQL-based systems.
- Work on backend features that need careful concurrency control and operational safety such as file-locking, atomic backups and transactional batch updates.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Senior Frontend Developer
Confidence: Medium Fullstack
Fullstack web engineer at a senior level with a practical focus on reliable backend data pipelines and pragmatic React frontends. The strongest proven skill is building safe, production data-processing pipelines and migration tooling as shown by Kanvance/backend/src/migrations/001_analytics_timesheet_refactor.js and Kanvance/backend/src/services/timesheetAutofill.service.js. Public code shows limited evidence for large-scale distributed system design, cloud-native infra automation, or specialized frontend performance telemetry such as bundle analysis and RUM pipelines.
UI Component Architecture
4/10
How interface parts are built
Component architecture shows custom hooks and a lightweight design system but limited evidence of a bespoke component library or advanced composition patterns across many components.
Evidence
yuvrajjaiswal.dev/src/hooks/use-toast.ts: custom global toast implementation with in-memory state and listener pattern
yuvrajjaiswal.dev/src/index.css: design tokens and global styling for a custom design system
yuvrajjaiswal.dev/src/hooks/use-reveal-on-scroll.ts: small reusable hook for intersection-based reveal
Responsive & Cross-browser
6/10
Works on all screens and browsers
Responsive design and cross-browser considerations are explicit, with detailed media queries, container sizing, and prefers-reduced-motion handling.
Evidence
yuvrajjaiswal.dev/src/index.css: media queries, prefers-reduced-motion and focus-visible rules
yuvrajjaiswal.dev/src/tailwind.config.ts: container and screen settings plus theme tokens
Kanvance/frontend/src/components/Projects/Projects.module.css: extensive responsive rules for multiple breakpoints and mobile adaptations
Performance Optimization
4/10
Speed of the interface
Some measured/performance-aware patterns exist such as IntersectionObserver for reveal, test performance expectations, and efficient batch processing on the backend, but no visible bundle-level analysis or advanced client rendering optimizations.
Evidence
yuvrajjaiswal.dev/src/hooks/use-reveal-on-scroll.ts: IntersectionObserver with reduced-motion guard
daily-activity-log-parser-for-excel/cyberark-timesheet/backend/tests/timesheetParser.test.js: performance test asserting 500+ rows parse time
Kanvance/backend/src/migrations/001_analytics_timesheet_refactor.js: migration executed in transactional batches and dry-run mode
Accessibility & Semantics
6/10
Usable for everyone
Accessibility and semantics are considered in CSS and hooks: focus-visible outlines, prefers-reduced-motion respect, expanded touch targets and keyboard focus styles are present.
Evidence
yuvrajjaiswal.dev/src/index.css: focus-visible styling and prefers-reduced-motion handling
Kanvance/frontend/src/components/Projects/Projects.module.css: keyboard focus selectors and expanded touch targets for interactive elements
yuvrajjaiswal.dev/src/hooks/use-reveal-on-scroll.ts: respects prefers-reduced-motion and avoids animation when reduced-motion is requested
State Management & Data Flow
5/10
Managing data in the app
State management shows deliberate choices: a small custom global state for toasts, Redux in a frontend app, and server-side orchestration with dry-run, idempotency and race handling; evidence of advanced client-server cache patterns is limited in the sampled files.
Evidence
yuvrajjaiswal.dev/src/hooks/use-toast.ts: custom pub-sub style toast state with dispatch and listeners
Kanvance/frontend/src/redux/store.js: Redux store wiring in the frontend
Kanvance/backend/src/services/timesheetAutofill.service.js: orchestrator using dry-run, reset-on-race, and careful upsert logic
UX & Visual Polish
6/10
Look and feel quality
Strong visual polish and UX attention are visible: design tokens, micro-interactions, skeleton and reveal patterns, and comprehensive styling for complex list and card UIs.
Evidence
yuvrajjaiswal.dev/src/index.css: design system tokens, animations, micro-interactions and hero timing
Kanvance/frontend/src/components/Projects/Projects.module.css: extensive polished styles for cards, responsive layout and accessible focus states
yuvrajjaiswal.dev/src/hooks/use-reveal-on-scroll.ts: reveal logic contributing to perceived performance and UX
Expertise
React• Middle
Frontend Architecture & Build Tools• Middle
Industries
Information Technology• Middle
Software• Middle
Technologies
JavaScript
TypeScript• since 2025 • Senior
Node JS• Senior
Tailwind CSS
Express
React.js
Vite
React Query
Axios
Zod
React Hook Form
Redux Toolkit
React Router
Recommendations
- Lead development of internal productivity and analytics tools that combine Excel/ETL pipelines with a React frontend and a Node API.
- Implement backend data-migration and safe-deploy features such as idempotent migrations, dry-run modes and atomic file operations for enterprise tooling.
- Build and maintain a component-driven design system and accessible UI library that the product teams can consume.
- Own feature work that requires careful client-server state discipline, validation and rollback patterns (e.g., timesheet import/enrich/confirm flows).
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: