Nikita Gorokhov
Backend Developer
9+ years exp
7+ years ML exp
10+ projects
SQL
Node JS
JavaScript
C++
PHP
C
Python
Java
Scala
TypeScript
Data Layer & Database: 8/10
API Design: 6/10
Scalability & Performance: 6/10
Active 14 days ago
Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
A seasoned engineering generalist focused on building robust local-first AI tooling and backend services at a senior level. The strongest proven skill is designing a production-focused persistent memory and search stack (SQLite + FTS5 + hybrid vector search) with idempotent migrations and comprehensive unit/integration tests - see mnemon-memory-mcp/src/db.ts and src/tools/memory-search.ts. There is limited public evidence of formal CI flakiness remediation, large-scale performance testing (load/soak), or an enterprise-grade flaky-test quarantine system.
Technical skills
SQL
Node JS• Senior
JavaScript
C++• Middle • 9y+
PHP• 9y+
C• Middle • 9y+
Python• Senior • 7y+
Java• Middle • 7y+
Scala• Middle • 6y+
TypeScript• Senior
Node JS
BullMQ
Prisma
Express
Nest.JS
Python
FastAPI
SQLAlchemy
Databases
ClickHouse
PostgreSQL
Redis
SQLite
MongoDB
Apache Kafka• 7y+
ElasticSearch• 7y+
CockroachDB• 6y+
Frontend
Next.js
React.js
Astro
Tailwind CSS
npm• 7y+
DevOps
AWS
Docker
GCP
GitHub Actions
Grafana
Prometheus
Terraform
Rest API
CI/CD
Helm
Kubernetes
QA
Sentry
Vitest
Playwright
Pytest
AI/ML
AI Agents
Claude
Cursor
Embeddings
Function Calling
Gemini
LLM
RAG
TensorFlow• 7y+
OpenAI SDK• 5+ projects
Claude Code• 10+ projects
Web3
DAO
Staking
Ethereum
Algorand• 5y+
DeFi• 5y+
NFT• 5y+
Polygon
The Graph
Timeline
Co-Founder, CEO & CTO
•
Executive
MetaLabs (cometa.farm)
•
Cofounder
Co-founded and built a DeFi platform on Algorand from the ground up, covering staking pools, a DEX aggregator, NFT collections, and a 3D play-to-earn game. Raised funding and led engineering execution, including hiring and managing a team up to ten people. Currently runs the system in maintenance mode with unified technical ownership across areas of the product.
Algorand
DeFi
NFT
Senior / Lead Backend Engineer
•
Lead
V-Boost (MetaWhale)
•
Contractor
Built the backend from scratch, including production NestJS/TypeScript services and a The Graph subgraph, plus additional services nearing launch. Architected an on-chain/off-chain subscription engine with reconciled payment state to prevent access drift. Focused on launch readiness and reliability with staged Helm rollouts, observability, and on-call response coverage for a 100k+ user system.
Nest.JS
TypeScript
The Graph
Polygon
Kubernetes
Helm
Senior Backend Engineer / Team Lead
•
Lead
the_covert
•
Full-Time
Led backend development for a B2B custodial crypto-asset platform with real-time transactions and trading capabilities. Integrated major exchanges and blockchain nodes, leveraging native C cryptography within a Scala/CockroachDB stack. Managed engineering direction and delivery for core trading and transaction flows.
Scala
CockroachDB
C
Backend Developer
•
Middle
Yandex
•
Full-Time
Shipped backend features for Yandex Tracker on a high-load Java/Spring platform. Worked with event/data infrastructure including Kafka, Elasticsearch, and MongoDB to support scale and reliability. Focused on performance and stability of backend services handling tracker workloads.
Java
Apache Kafka
ElasticSearch
Software Engineer
•
Middle
Assaia
•
Full-Time
Developed AI video event-recognition for airport operations, building the data and training pipeline around model development. Implemented supporting tooling for preparing inputs and improving training workflows. Used Python and TensorFlow to support end-to-end production model delivery.
Python
TensorFlow
Software Engineer
•
Middle
VK
•
Full-Time
Worked on core C/C++ backend infrastructure for a high-scale social network serving hundreds of millions of users. Built a k-NN HNSW recommendation engine and contributed to KPHP, VK’s PHP compiler. Focused on performance-critical systems and core infrastructure reliability.
Csince 2017
C++
PHP
SWE Intern
•
Junior
Google
•
Part-Time
Added quota and throttling logic to a core network library handling very high request throughput. Supported reliability and stability of request processing under load conditions. Contributed to network-layer protections during internship.
SRE Intern
•
Junior
Google
•
Contractor
C++since 2017
Javasince 2017
Senior Backend Developer
Confidence: Medium Data Platform
A backend engineer at a senior level focused on local-first agent memory and retrieval systems. The strongest proven skill is data-layer and search engineering - evidenced by a multi-version SQLite migration chain, FTS5 trigger tuning and a hybrid FTS+vector search implementation (src/db.ts and src/tools/memory-search.ts). There is limited public evidence for multi-node distributed architecture, formal API versioning, or long-lived token lifecycle management.
API Design
6/10
How well APIs are designed
Solid API/tooling design for an MCP-based server with input schema validation, consistent error redaction, pagination and filtering; lacks a formal public versioning strategy or documented idempotency keys on HTTP endpoints.
Evidence
nikitacometa/mnemon-memory-mcp/src/server.ts: createMcpServer - tool registration, schema validation and CallToolRequestSchema handler
nikitacometa/mnemon-memory-mcp/src/tools/memory-search.ts: pagination (limit/offset), has_more and query routing for modes
nikitacometa/mnemon-memory-mcp/src/http-server.ts: isRateLimited / authorization helpers (HTTP boundary handling)
Data Layer & Database
8/10
Working with databases
Clear, careful data-layer engineering: multi-version migrations, idempotent migration helpers, FTS5 triggers, partial indexes, index-time stemming/backfill and a vector-index meta table - all using transactions and prepared statements.
Evidence
nikitacometa/mnemon-memory-mcp/src/db.ts: runMigrations + applyMigration1..applyMigration8, safeAddColumn/safeDropColumn, FTS5 triggers and backfillStemmedContent
nikitacometa/mnemon-memory-mcp/src/tools/memory-search.ts: use of prepared SQL, field-weighted bm25() query and filter construction
nikitacometa/mnemon-memory-mcp/src/vector.ts: vector_index_meta creation and sqlite-vec integration points
Scalability & Performance
6/10
Handling load and speed
Performance-conscious design with WAL mode, busy_timeout, FTS tuning and hybrid vector+FTS search with progressive KNN expansion and over-fetch to maintain correct pagination; no explicit distributed-scaling or multi-node strategies (expected for a local-first SQLite service).
Evidence
nikitacometa/mnemon-memory-mcp/src/db.ts: db.pragma('journal_mode = WAL'), db.pragma('busy_timeout = 5000')
nikitacometa/mnemon-memory-mcp/src/tools/memory-search.ts: hybrid/vectorSearch + knnLimit expansion and over-fetch/fetchLimit logic
nikitacometa/mnemon-memory-mcp/src/tools/memory-search.ts: bm25 field weights and AND/OR relaxation to balance recall/precision
System Architecture
6/10
Overall system structure
Reasonable module decomposition and separation of concerns: a central MCP server factory, transport-specific entrypoints (stdio vs HTTP), import pipeline and vector/fts/search modules; architecture favors a single-node local-first pattern rather than microservice distribution.
Evidence
nikitacometa/mnemon-memory-mcp/src/server.ts: central createMcpServer, tool/resource/prompt registration and clear transport separation
nikitacometa/mnemon-memory-mcp/src/index.ts / src/index-http.ts: transport entrypoints (stdio vs HTTP)
nikitacometa/mnemon-memory-mcp/src/import/kb-import.ts: dedicated import/ETL pipeline with idempotency and source_file superseding
Security & Auth
6/10
Protecting data and access
Good input-safety and secrets hygiene at the app level - runtime permission hardening, zod-based validation, prepared statements and sanitized error messages; no broad OAuth/token lifecycle artifacts or revocation flows present in the analyzed code.
Evidence
nikitacometa/mnemon-memory-mcp/src/db.ts: directory/file permission hardening (mkdirSync + chmodSync) and WAL sidecar permission handling
nikitacometa/mnemon-memory-mcp/src/server.ts: safe error redaction before returning tool errors to clients
nikitacometa/mnemon-memory-mcp/src/validation.ts and server.ts: Zod schemas used to parse/validate tool inputs
Reliability & Observability
6/10
Stability and monitoring
Reliability and observability are thoughtfully addressed - migrations are transactional and idempotent, search queries are logged (best-effort) with pruning, and critical failures are captured in structured stderr logs; there is limited evidence of multi-process graceful shutdown orchestration or distributed traces/alerts.
Evidence
nikitacometa/mnemon-memory-mcp/src/db.ts: transactional migrations and backfill logic (db.transaction wrappers)
nikitacometa/mnemon-memory-mcp/src/tools/memory-search.ts: logSearch() persists search_log with pruning and is best-effort
nikitacometa/mnemon-memory-mcp/src/server.ts: structured re-embedding failure logs written to stderr and try/catch boundaries for tool execution
Verified artifacts
Expertise
Backend AI & LLM• Middle
Databases & Vector Storage• Senior
Microservices & API Architecture• Middle
Technologies
MongoDB
SQLite
Cloud• mentioned only
Persistent• mentioned only
Vector• mentioned only
Recommendations
- Lead development of embedding-backed retrieval services and local-first memory stores (indexing, FTS tuning, sqlite-vec integration).
- Build retrieval and search features for AI assistants - hybrid FTS+vector ranking, RRF fusion and snippet generation.
- Implement data-layer work such as schema evolution, migration safety, and import/ETL pipelines for document/knowledge ingestion.
- Harden operational observability - structured tracing, SLO-driven metrics and automated reindex/backup procedures for production deployments.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Senior QA Engineer
Confidence: High Generalist
A seasoned engineering generalist focused on building robust local-first AI tooling and backend services at a senior level. The strongest proven skill is designing a production-focused persistent memory and search stack (SQLite + FTS5 + hybrid vector search) with idempotent migrations and comprehensive unit/integration tests - see mnemon-memory-mcp/src/db.ts and src/tools/memory-search.ts. There is limited public evidence of formal CI flakiness remediation, large-scale performance testing (load/soak), or an enterprise-grade flaky-test quarantine system.
Test Automation Frameworks
6/10
Building automated tests
Well-instrumented unit/integration test runner (Vitest) and many focused tests exist; however there is limited evidence of a separate, large-scale bespoke test framework (fixtures, testcontainers, or sophisticated sharding).
Evidence
mnemon-memory-mcp/package.json: 'test': 'vitest run'
mnemon-memory-mcp/vitest.config.ts: test include pattern 'src/**/__tests__/**/*.test.ts'
mnemon-memory-mcp/src/__tests__/server.test.ts
Test Coverage & Strategy
6/10
What and how to test
Code shows deliberate negative-path and boundary handling (schema migrations, idempotent operations, explicit storage error types) and focused integration tests; mutation testing or property-based tests are not evident.
Evidence
mnemon-memory-mcp/src/db.ts: safeAddColumn/safeDropColumn and migration idempotency (runMigrations, backfillStemmedContent)
mnemon-memory-mcp/src/tools/memory-search.ts: MemorySearchStorageError and isFts5QuerySyntaxError explicit handling
mnemon-memory-mcp/src/tools/__tests__/integration.test.ts
API & Integration Testing
6/10
Testing how parts work together
API and integration testing is present with explicit HTTP handlers, request validation and server tests; there is good coverage of API error handling but fewer traces of contract-testing tooling (pact/schemathesis) or containerized integration harnesses.
Evidence
mnemon-memory-mcp/src/http-server.ts: createHttpServer and request handlers with auth/rate handling
mnemon-memory-mcp/src/__tests__/http-server.test.ts
mnemon-memory-mcp/src/tools/__tests__/integration.test.ts
Performance & Load Testing
4/10
Testing speed under load
Performance considerations are visible (benchmarks, over-fetching and adaptive knn expansion, constants for KNN control), but there is no evidence of load/soak tooling, SLO-driven thresholds, or systematic percentile analysis.
Evidence
mnemon-memory-mcp/src/__benchmarks__/search.bench.ts
mnemon-memory-mcp/src/tools/memory-search.ts: VECTOR_KNN_INITIAL_MULTIPLIER, VECTOR_KNN_EXPANSION_FACTOR and adaptive knn expansion logic
Bug Reporting & Analysis
3/10
Finding and describing bugs
Project includes issue templates and contributing docs and many descriptive test failures, but there are no clearly linked public bug reports with reproductions and root-cause commits in the analyzed artifact set.
Evidence
mnemon-memory-mcp/.github/ISSUE_TEMPLATE/bug_report.md
mnemon-memory-mcp/CONTRIBUTING.md
CI Test Integration
5/10
Running tests automatically
CI/test integration artifacts exist (test scripts, vitest config and a Playwright e2e configuration with retries/tracing), but I see no fully parameterized CI matrix, per-test quarantine mechanics, or long-term flaky-test telemetry in the analyzed files.
Evidence
fairground/e2e/playwright.config.ts: retries:1, trace: 'on-first-retry', screenshot: 'only-on-failure'
mnemon-memory-mcp/package.json: npm test/vitest scripts
mnemon-memory-mcp/vitest.config.ts
Expertise
API Testing & Contract Validation• Middle
SDET & Test Engineering• Middle
Unit & Component Testing• Middle
Industries
Artificial Intelligence• Senior
Blockchain & Crypto• Middle
Gaming• Middle
Technologies
CI/CD
Python• Senior • 7y+
TypeScript• Senior
Playwright
Pytest
Sentry
Vitest
Cloud• mentioned only
Vector• mentioned only
Recommendations
- Lead development of local-first AI memory and retrieval services (hybrid FTS + vector search, embedding pipelines, idempotent migrations).
- Design and own backend tooling integrating MCP/JSON-RPC servers, embedding providers and re-embedding flows (embedder integration and vector index lifecycle).
- Implement and harden API/integration test suites and contract tests (HTTP + MQTT/MCP) with CI artifact capture and per-test quarantine for flaky cases.
- Tackle Algorand/crypto backend components where high-assurance tests and Python-based contract tests are needed (smart-contract test harnesses and property checks).
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories:
Middle Frontend Developer
Confidence: Low Fullstack
A senior-level fullstack engineer who designs and implements robust, production-grade backend and fullstack systems with strong emphasis on data integrity and search/embedding infrastructure. The strongest proven skill is building reliable, performant memory/search systems with concrete artifacts in mnemon-mcp (db migrations, FTS5 + stemmed indexing, hybrid FTS/vector RRF fusion in src/tools/memory-search.ts). Public code shows limited evidence of a bespoke UI design system, advanced accessibility engineering, or detailed frontend component libraries.
UI Component Architecture
3/10
How interface parts are built
Limited evidence of intentional UI component architecture - some custom components and UI generation exist but there is no clear, original design-system or detailed component boundary rationale visible in the human-authored files.
Responsive & Cross-browser
2/10
Works on all screens and browsers
Basic responsive work is present (component layouts and global CSS), but there is little sign of advanced cross-browser, container-queries, RTL or progressive enhancement engineering in the inspected human-authored UI files.
Performance Optimization
6/10
Speed of the interface
Strong, practical performance engineering on backend/search and I/O - careful DB pragmas, over-fetching logic to compensate for client-side re-ranking, iterative KNN expansion and best-effort logging; measurable optimization artifacts and defensive design exist but no published before/after metrics were found in code.
Accessibility & Semantics
2/10
Usable for everyone
Some attention to safe text encoding and escaping is visible for bot/UI outputs, but there is minimal evidence of systematic accessibility work (focus management, ARIA on custom widgets, CI axe checks) in human-authored UI code.
State Management & Data Flow
7/10
Managing data in the app
Strong server- and data-state discipline: idempotent migrations, transactional updates, explicit invalidation/re-embedding flows, guarded import pipeline, and hybrid search orchestration (FTS + vector) with graceful degradation and typed storage errors.
Evidence
UX & Visual Polish
5/10
Look and feel quality
Good UX attention in conversational flows - explicit loading/wait patterns, clear error handling and recovery in bot routes, and many test cases covering edge/error states; less evidence of UI-level perceived-performance or advanced skeletons for web frontends.
Evidence
nikitacometa/euphoria-ai/src/bot/routes/journal-entry.ts - withWaitMessage usage and multi-path error handling
nikitacometa/euphoria-ai/src/ai/journal-ai.test.ts - tests for graceful API failures
nikitacometa/mnemon-memory-mcp/src/tools/memory-search.ts - query_time_ms and has_more pagination handling
Expertise
React• Middle
Modern Web Frameworks• Middle
Web Performance & Optimization• Middle
Industries
Artificial Intelligence• Senior
Blockchain & Crypto• Senior
Gaming• Middle
Technologies
JavaScript
Node JS• Senior
Nest.JS
Tailwind CSS
Next.js
Express
Prisma
React.js
npm• 7y+
BullMQ
Astro
Cloud• mentioned only
Vector• mentioned only
Recommendations
- Lead or implement backend-centered features requiring reliable search, ranking and vector/embedding integrations (design and harden hybrid search pipelines and migration strategies).
- Build server-side components and BFFs for AI-enabled features - memory/context services, tool integrations, and re-embedding/invalidations.
- Develop fullstack Web3 game features that require strong server-side correctness and client recovery flows (on-chain verification, session recovery, proof/payout pipelines).
- Implement conversational UX and resilient media-processing flows for chatbots or assistant integrations (transcription, structured LLM calls, graceful error paths).
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: