Kirill Ismanov
Analyst
Management: 1-2 years (1-5 people)
SQL
Python
Active 2 days ago
Invite to interview
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
Data Analyst since 2022 with experience in fintech, B2B sales, and e-commerce. Skilled in SQL, Python, Excel, and BI tools. I analyze product and business metrics, investigate performance changes, and build dashboards to support decisions. Comfortable building analytics from scratch and working independently. My previous experience in sales management helps me understand business needs and translate data into practical insights.
Phone
Technical skills
SQL
Python
Python
SQLAlchemy
AI/ML
Jupyter Notebook
ChatGPT
Claude
Airflow
Analytics
QlikSense
Microsoft Excel
ETL/ELT
Databases
Greenplum
Teradata
DevOps
Jenkins
GitLab
Timeline
Product Analyst
•
Middle
TELE2 Russia
•
Full-Time
Joined the product launch for the new TELE2 card and payments ecosystem and owned analytics for multiple directions after the team lead left. Built and maintained dashboards for card openings, operations, balances, and mobile commerce performance, including segmentation by activity and regions. Investigated operational errors, identified signals of potential fraud, and helped correct partner contract logic by initiating an amendment. Automated mobile commerce reporting and delivered daily/recurring insights using data-mart updates scheduled with Airflow.
SQL
Teradata
Python
Airflow
Microsoft Excel
GitLab
Data Analyst
•
Middle
Komus
•
Full-Time
Supported B2B sales analytics across clients, revenue, product categories, and delivery performance, covering both regular reporting and ad hoc analyses. Created a new management report from multiple sources (about 50 KPIs), including client risk/rule monitoring and metrics like penetration and segment-based dynamics. Automated and modernized reporting from legacy Luigi/Jenkins solutions to new outputs, assisting colleagues with Python where needed.
SQL
Python
Microsoft Excel
Jenkins
Technical Support Analyst
•
Middle
Sber
•
Full-Time
Coordinated technical support analytics for a portfolio of bank products and then became the primary analyst for department reporting. Improved SLA from 89% to 98% by identifying underperforming products and analyzing request types to distinguish between bugs, consulting needs, and missing documentation. Migrated manual Excel reporting to automated Excel workflows and Qlik Sense dashboards, produced SLA/cost/penalty metrics, and supported onboarding of teams with reporting methodology.
Microsoft Excel
QlikSense
Python
pySpark
Middle Backend Developer
Confidence: Medium Data Platform
Middle backend engineer specializing in LLM-driven analytics and SQL automation. The strongest proven skill is integrating LLM clients with data pipelines, evidenced by srs/ai/llm.py GigaClient/OllamaClient and the orchestration in srs/main.py. There is little evidence of production-grade API design, robust security controls for executing generated SQL, migration history, or observability and operational tooling.
API Design
2/10
How well APIs are designed
Basic integration wrappers for LLM clients and explicit system prompts exist, but there is no formal API contract, versioning, pagination, or idempotency handling.
Evidence
srs/ai/llm.py: OllamaClient.ask and GigaClient.ask wrapper implementations
srs/ai/prompts.py: SQL_GENERATOR_PROMPT system prompt that constrains LLM output
Data Layer & Database
3/10
Working with databases
Shows schema introspection and transactional bulk loading, but lacks migration history, tuned queries, and explicit transaction/isolation-level handling beyond simple use of engine.begin.
Evidence
srs/db/schema.py: get_columns queries information_schema and returns DataFrame
srs/etl/load_data.py: uses engine.begin() to load multiple tables in a transaction
Scalability & Performance
1/10
Handling load and speed
No evidence of caching strategy, queue-based decoupling, rate limiting, or measured performance tuning; single-process ETL and direct DB calls are used.
Evidence
srs/etl/load_data.py: direct CSV-to-DB loads using write_dataframe with engine.begin()
System Architecture
3/10
Overall system structure
Reasonable separation into ai, db, etl modules and a simple orchestrator in main.py, but the architecture is small-scale and lacks service contracts, graceful degradation patterns, or config/secret management beyond dotenv.
Evidence
srs/main.py: orchestration flow wiring LLM prompts, SQL generation and query execution
Project layout: srs/ai, srs/db, srs/etl folders indicate separation of concerns
Security & Auth
2/10
Protecting data and access
Environment-based secret handling is used, but executing arbitrary SQL produced by an LLM without sanitization and a default verify_ssl_certs=False are notable security gaps.
Evidence
srs/main.py: load_dotenv() and direct execution of LLM-generated SQL via execute_query
srs/ai/llm.py: GigaClient accepts credentials from os.getenv and sets verify_ssl_certs=False by default
Reliability & Observability
2/10
Stability and monitoring
Minimal reliability and observability patterns are present such as a retry path for failed SQL and token usage printouts, but there are no structured logs, metrics, timeouts, backoff, or graceful shutdown handling.
Evidence
srs/main.py: try/except around execute_query with a retry prompt path
srs/ai/llm.py: print of response.usage.total_tokens for token visibility
Expertise
Backend AI & LLM• Middle
Databases & Vector Storage• Middle
Python• Middle
Industries
Artificial Intelligence• Middle
Data & Analytics• Middle
Technologies
Python• since 2024 • Middle
SQLAlchemy
Recommendations
- Develop LLM-driven analytics prototypes and internal tools that translate natural language to SQL while keeping them in an isolated sandbox for testing.
- Focus on safe SQL execution patterns - implement execution sandboxes, query timeouts, result row limits, and strict validation of generated SQL before running in production.
- Build more robust ETL and schema evolution practices - add migration history, idempotent loaders, and explicit transaction/isolation strategies.
- Harden integrations with proper secret management, SSL validation defaults, structured logging, and instrumentation (metrics and traces) for observability.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: