Ramazan Jalalidinov
Python Developer
4+ years exp
Bash
SQL
Node JS
TypeScript
Python
JavaScript
Data Pipeline & Feature Engineering: 4/10
Active 19 days ago
+996 (558) 888245 Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
A practical engineer focused on building production tooling for reporting and transcription workflows at a mid-senior level. The strongest proven skill is engineering robust data+inference pipelines - evidenced by the ReportGenerator (reports_maker/apps/ipi/report_generator.py) and the LLM-backed diarization postprocessing (transcription-tool/llm_speaker_correct.py). There is little to no evidence of custom model training, experiment tracking, or large-scale MLOps (model versioning, drift monitoring) in the public code.
Technical skills
Bash
SQL
Node JS
TypeScript
Python• Senior • 4y+
JavaScript• Junior • 4y+
Node JS
Nest.JS
Python
SQLAlchemy
Requests
Celery• 3y+
Django• 3y+
FastAPI• 3y+
Flask• 3y+
Databases
ClickHouse
Databases
PostgreSQL• 3y+
Redis• 3y+
MS SQL
AI/ML
Pandas
NumPy
OpenAI SDK
Speech Recognition
LangChain
LLM
Streamlit
Frontend
Bootstrap
JQuery
npm
DevOps
Bitbucket
Azure
Docker• 3y+
Git• 3y+
Rest API• 3y+
CI/CD
QA
Pytest
Analytics
Power BI
Timeline
AI Engineer
•
Middle
adastra-inc
•
Full-Time
Built automation and data processing workflows, including web automation, data validation, reporting, and LLM-assisted tools. Developed data pipelines and integrations to process business data and generate operational reports. Implemented internal reporting applications using Python and Microsoft SQL Server with Power BI, and created LangChain-based MCP integrations for secure LLM access to internal systems.
Python
MS SQL
Power BI
Streamlit
Django
FastAPI
CI/CD
LangChain
LLM
Docker
Python Backend Developer
•
Middle
Transparency
•
Full-Time
Developed and maintained backend services for multiple web platforms, implementing core backend functionality and supporting REST API endpoints. Built database models and backend components using Django with PostgreSQL. Took part in architecture discussions, debugging, code reviews, and also contributed to a browser game using web technologies.
Pythonsince 2023
Djangosince 2023
Rest API
FastAPIsince 2023
Flask
PostgreSQL
Redis
Celery
Dockersince 2023
Git
Senior AI/ML Engineer
Confidence: High LLM Engineer
A practical engineer focused on building production tooling for reporting and transcription workflows at a mid-senior level. The strongest proven skill is engineering robust data+inference pipelines - evidenced by the ReportGenerator (reports_maker/apps/ipi/report_generator.py) and the LLM-backed diarization postprocessing (transcription-tool/llm_speaker_correct.py). There is little to no evidence of custom model training, experiment tracking, or large-scale MLOps (model versioning, drift monitoring) in the public code.
Model Architecture & Training
2/10
How well models are designed and trained
No model training code or custom architectures - LLM usage is limited to inference and guarded prompt handling. The project includes careful LLM prompting and postprocessing but not training/architecture work.
Evidence
transcription-tool/llm_speaker_correct.py:postprocess_with_llm
transcription-tool/llm_speaker_correct.py:apply_corrections
Data Pipeline & Feature Engineering
4/10
How data is prepared for models
Solid data engineering for reporting and transcription: DataFrame normalization, single-query DB fetch optimization, segmentation for audio results and CSV/Excel/PDF generation pipelines.
Evidence
reports_maker/apps/ipi/report_generator.py:get_dataset
reports_maker/apps/ipi/report_generator/report_generator.py:ReportGenerator.main
transcription-tool/assemblyai_transcribe.py:segments_from_words
Experimentation & Evaluation
2/10
How results are measured and tested
Some evaluation utilities and heuristics for diarization quality but no formal experiment tracking or A/B pipelines.
Evidence
transcription-tool/diarization_eval.py:speaker_accuracy
transcription-tool/diarization_eval.py:evaluate_variant
MLOps & Deployment
3/10
How models are shipped to production
Application-level serving and deployment artifacts exist (Streamlit app, Selenium-based PDF generation, robust API polling/retries) but there is no full model lifecycle/versioning or drift monitoring.
Evidence
transcription-tool/streamlit_app.py:Streamlit UI + user flows
reports_maker/apps/ipi/report_generator.py:PdfGenerator (Selenium + ChromeDriverManager)
transcription-tool/assemblyai_transcribe.py:_api_request (retry/backoff)
Computational Efficiency
3/10
How efficiently computing resources are used
Practical efficiency work: minimizing DB roundtrips, streaming downloads, and retry/backoff logic. No GPU/quantization or advanced performance profiling evidence.
Evidence
reports_maker/apps/ipi/report_generator/report_generator.py - single DB fetch optimization (load once, filter later)
transcription-tool/plunet_order_client.py:ProgressTransport streaming download
transcription-tool/assemblyai_transcribe.py:_api_request (retry/backoff)
Research Depth & Innovation
2/10
Depth of research and new ideas
Applied heuristics and conservative safety rules for LLM-based speaker correction and segmentation; useful and pragmatic but not research-level novel algorithm work.
Evidence
transcription-tool/llm_speaker_correct.py:_reject_correction_reason (safety heuristics)
transcription-tool/assemblyai_transcribe.py:segments_from_words (segmentation heuristics)
Expertise
Audio & Speech Processing• Middle
LLM• Middle
Technologies
Databases
Python• Senior • 4y+
SQL
PostgreSQL• 3y+
Redis• 3y+
LangChain
ClickHouse
Flask• 3y+
FastAPI• 3y+
Speech Recognition
OpenAI SDK
Azure
CI/CD
Pandas
NumPy
Django• 3y+
Git• 3y+
Docker• 3y+
LLM
Celery• 3y+
Streamlit
Requests
Bitbucket
Recommendations
- Lead development of LLM-assisted transcription and QA tools - extend the llm_speaker_correct pipeline into a deployable microservice with batching, retries and auditing.
- Build and harden data pipelines and reporting backends - productionize the ReportGenerator with CI/CD, secrets management, and parametric job orchestration (e.g., Airflow/Argo).
- Implement experiment tracking and evaluation - add ML experiment logging (W&B/MLflow), automated diarization evaluation, and datasets/holdouts for reproducible improvements.
- Harden security and secrets handling - remove hardcoded credentials, adopt environment/secret store, and tighten Selenium flags and file access policies before deployment.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Senior Backend Developer
Confidence: Medium API Engineer
A pragmatic backend engineer at a mid-senior level who builds end-to-end Django services, data pipelines and integrations. The strongest proven skill is building data-heavy report and ETL-style systems with concrete artifacts - notably the ReportGenerator/report_builder/report_generator.py pipeline and its single-query DB fetch + pandas normalization. What is not evidenced publicly is formal ops (SLI/SLO/alerting), extensive automated tests, or production-grade secret management.
API Design
5/10
How well APIs are designed
Practical, consistent REST endpoints with structured JSON responses, use of decorators (@login_required, @require_http_methods), and custom JWT auth/middleware show API-focused design; lacks explicit versioning, formal error contract schema, idempotency keys or pagination strategy.
Evidence
reports_maker/apps/ipi/views.py: save_custom_template_set / generate_report / get_report_history (JSON endpoints, status codes)
reports_maker/apps/user/backends.py: JWTAdminBackend (custom JWT auth backend)
reports_maker/apps/user/middleware.py: JWTAuthMiddleware (token parsing on request)
Data Layer & Database
5/10
Working with databases
Shows deliberate data-layer work: single large SQL query to avoid N+1, normalization pipeline with pandas, and a real migration history; however there is limited evidence of transaction boundary handling or explicit isolation-level controls.
Evidence
reports_maker/apps/ipi/report_generator.py: single combined SQL query and get_interpreting_data usage to fetch all rows once
reports_maker/apps/ipi/report_generator/database.py: get_db_connection / normalize_df (pyodbc + pandas normalization)
reports_maker/apps/ipi/migrations (0001_initial..0006_rename_account_to_accountgroup): multiple migration files showing schema evolution
Scalability & Performance
4/10
Handling load and speed
Performance-minded choices are present (one-shot DB fetch, bulk_create for seeds, reusable dataframes), and HTTP/soap clients use retries and timeouts; there is limited evidence of system-level caching/invalidation, connection pooling configuration, or measured load-testing artifacts.
Evidence
reports_maker/apps/ipi/report_generator.py: fetch all date ranges / single DB query to minimize round-trips
transcription-tool/plunet_order_client.py: _make_session with Retry and HTTPAdapter (retries, pool settings)
turizm/heatmap/management/commands/seed_demo.py and seed_100.py: bulk_create usage and synthetic workload seeding
System Architecture
5/10
Overall system structure
Clear modular decomposition (report_generator split into database, builder, pdf_generator, charts; separate transcription tool modules), a reusable ReportGenerator class and recursive report-folder strategy; overall a monolith Django deployment rather than distributed microservices.
Evidence
reports_maker/apps/ipi/report_generator/report_generator.py: ReportGenerator class with recursive generate_report and clear module boundaries
reports_maker/apps/ipi/report_generator/report_builder.py and pdf_generator.py: separation of concerns (prepare/build PDF/HTML/Excel)
transcription-tool/*: modular layout (assemblyai_transcribe, llm_speaker_correct, streamlit_app)
Security & Auth
3/10
Protecting data and access
Some security-aware components exist (JWT auth backend, token middleware), but there are clear secrets hygiene issues (hardcoded DB credentials) and several endpoints use csrf_exempt; input validation and explicit defenses (SQL injection, SSRF) are limited or not documented.
Evidence
reports_maker/apps/user/backends.py and middleware.py: custom JWT authentication and middleware
reports_maker/apps/ipi/report_generator.py: DB credentials hardcoded (DB_SERVER/DB_USERNAME/DB_PASSWORD)
reports_maker/apps/ipi/views.py: use of @csrf_exempt on some POST endpoints
Reliability & Observability
5/10
Stability and monitoring
Good observability and reliability practices: structured logging, custom session log handler to stream generator logs to UI, retries/backoff in external API clients, and careful cleanup in finally blocks; lacks documented SLIs/SLOs or automated alerting configs.
Evidence
reports_maker/apps/ipi/views.py: SessionLogHandler and setup_session_logger to collect/generate per-session logs
transcription-tool/assemblyai_transcribe.py: _api_request with retry/backoff and poll loop with timeout
transcription-tool/plunet_order_client.py: _call wrapper with retry logic and re-login on session expiry
Expertise
Backend AI & LLM• Middle
Databases & Vector Storage• Middle
Microservices & API Architecture• Middle
Python• Middle
Industries
Artificial Intelligence• Middle
Data & Analytics• Middle
Travel & Tourism• Middle
Technologies
Rest API• 3y+
SQLAlchemy
MS SQL
Recommendations
- Develop backend APIs and data-reporting pipelines that produce scheduled or on-demand PDFs/archives (use ReportGenerator and report_builder patterns).
- Implement transcription and LLM post-processing integrations (AssemblyAI + Azure OpenAI) including safe prompts and conservative correction logic (extend llm_speaker_correct).
- Build data normalization and analytics pipelines using pandas and SQL Server retrieval, plus robust migration management and schema evolution.
- Lead mid-sized feature work touching auth, sessioned logging and observability (improve JWT flows, session log handler, and structured logs/metrics integration).
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Junior Frontend Developer
Confidence: Medium UI Engineer
A frontend-focused developer at a junior-to-middle level who builds server-rendered web UIs and polished responsive pages. The strongest proven skill is HTML/CSS-driven UI implementation, supported by original templates and a tailored stylesheet (Instagram_clone/templates and static/assets1/style.css). There is limited evidence of advanced frontend engineering - little client-side state management, performance instrumentation, or test coverage is present in the developer-owned files.
UI Component Architecture
3/10
How interface parts are built
Component structure is template-driven with some modular includes (header, post blocks) but there is no evidence of a component library, design-system package or advanced composition patterns.
Evidence
Instagram_clone/templates/include/header.html: modular header include with search, theme toggle and profile button
Instagram_clone/templates/index.html: repeated post card structure and story/post blocks (server-side composition)
Responsive & Cross-browser
5/10
Works on all screens and browsers
Responsive layouts and breakpoints are implemented with handcrafted CSS, media queries and a desktop/mobile layout, showing deliberate cross-breakpoint behavior.
Evidence
Instagram_clone/static/assets1/style.css: multiple @media rules (min-width: 620px, 768px, 1024px) and responsive layout adjustments
Instagram_clone/templates/index.html: markup uses responsive patterns (stories, posts, side-menu) tied to CSS rules
Performance Optimization
1/10
Speed of the interface
Little to no evidence of measured performance work, code-splitting, bundle analysis or advanced runtime optimization - mostly static assets and vendor libraries.
Evidence
Instagram_clone/static/assets1/style.css: handcrafted CSS but no perf artifacts (bundle reports, lazy-loading strategy)
Instagram_clone/static/js/bootstrap.js: vendor bootstrap present (UI), not a performance optimization artifact
Accessibility & Semantics
2/10
Usable for everyone
Some basic semantic markup and visual focus styles appear, but there is limited systematic ARIA usage or accessibility tooling shown.
Evidence
Instagram_clone/templates/index.html: uses <img> with alt and semantic header structures
Instagram_clone/templates/include/header.html: uses button elements and title attributes for interactive controls
State Management & Data Flow
2/10
Managing data in the app
State is primarily server-driven (Django forms and POST flows); minimal client-side state management or advanced async handling is present in original code.
Evidence
Instagram_clone/templates/index.html: multiple POST forms with {% csrf_token %} for likes/saves/comments (server-driven flows)
Instagram_clone/templates/post_detail_view.html: comment form and server-rendered comments list
UX & Visual Polish
4/10
Look and feel quality
Clear visual polish and UX attention in CSS (design tokens, consistent spacing, responsive story and post UI), but lacks advanced UX patterns like skeleton loading, optimistic UI or undo flows.
Evidence
Instagram_clone/static/assets1/style.css: variables, component styles, polished post/story UI and responsive adjustments
Instagram_clone/templates/index.html: UX patterns for stories, posts, like/save interactions and suggestion side-panel
Expertise
HTML & CSS• Junior
Technologies
JavaScript• Junior • 4y+
TypeScript
Node JS
Nest.JS
Bootstrap
JQuery
npm
Recommendations
- Build and maintain marketing or consumer-facing sites and UI components where server-rendered templates are acceptable (Django-style SSR).
- Develop small to medium interactive widgets and landing pages that require solid responsive CSS and accessible markup (story/post UI, header, search, profile widgets).
- Work on consolidating a component library and design tokens (extract repetitive markup into includes/components and publish a small style system).
- Take ownership of frontend polish and accessibility fixes - run axe/Pa11y audits and add keyboard/focus management to interactive widgets.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: