Portfolio

TusharSharma

TS_27X

Software Engineer

About

CORE_ENGINE

01_PROFILE.MD

COMPILING SYSTEM ARCHITECTURE... 100%

DECODING COMPLEXITY.
SCALING SOLUTIONS.

> I am a Software Engineer who views code as a tool for architectural precision. Focusing on the intersection of high-performance systems and scalable user interfaces. Most recently owned the KPI analytics platform end-to-end at Salescode.ai.

Core_Stack

BACKEND

PythonFastAPIJavaNetty

DATA

PostgreSQLElasticsearchRedisKafka

AI / LLM

RAGEmbeddingsPrompt EvalClaude

$ cat /etc/skills/*.yml

Process

CURRENT_OPS

// previously deployed at Salescode.ai as a Software Engineer. now building Python/FastAPI services and LLM systems (RAG, multi-provider failover) over PostgreSQL, Redis, Elasticsearch and Kafka.

TECH.EXE

04_SYSTEM_DEPENDENCIES

$ cat /etc/skills/*.yml

languages.yml

# languages.yml

languages:

-Python

-JavaScript

-TypeScript

-Java

-Go

-SQL

backend_apis.yml

# backend_apis.yml

backend_apis:

-FastAPI (async)

-REST API Design

-SQLAlchemy

-OAuth2/JWT

-Spring Boot

-Netty

-Microservices

data_messaging_cloud.yml

# data_messaging_cloud.yml

data_messaging_cloud:

-PostgreSQL

-MySQL

-Redis

-Elasticsearch

-Kafka

-GraphQL

-AWS (EC2, S3, DynamoDB)

ai_llm_systems.yml

# ai_llm_systems.yml

ai_llm_systems:

-RAG

-Vector Search

-Prompt Design & Eval

-LLM Orchestration

-Multi-provider Failover

-Structured Outputs

devops_practices.yml

# devops_practices.yml

devops_practices:

-Jenkins CI/CD

-Docker

-Git

-Logging & Monitoring

-Root-cause Analysis

frontend.yml

# frontend.yml

frontend:

-React.js

-Next.js

-Tailwind CSS

ALL_SYSTEMS_GO // HASH: 0xTS27A

Journey

HISTORY_LOG

02_EXPERIENCE.EXE

PARSING DEPLOYMENT HISTORY... SUCCESS

...

Aug 2024 - Aug 2026

Salescode.ai

Aug 2024 — Aug 2026

Salescode.ai

01011001 01011001 01011001 01011001 01011001

Software Engineer

// Backend Engineer for SalesLens (KPI analytics & dashboard platform). Owned the KPI analytics service used by 50-60 clients end-to-end — from gathering requirements to production support — across Elasticsearch, Redis, Kafka and AWS.

Cut dashboard API latency from 700-800ms to under 200ms by building a runtime query aggregation API with Redis caching, replacing a static layout that re-aggregated 2 years of history on every request.

Enabled real-time KPI breakdowns by user, location and division under 300ms with a pivot index strategy on Elasticsearch.

Reduced an ETL pipeline's runtime by 90% (30-40 minutes to 3-5) by rebuilding how it consumed ML-generated recommendations from AWS S3.

Eliminated batch delay in KPI reporting with an event-driven scheduler polling every 30 seconds and orchestrating in-memory KPI calculation.

Automated multi-step KPI batch runs on Jenkins CI/CD, turning hand-started runs with no recovery path into scheduled, repeatable, recoverable ones.

Onboarded and supported 50-60 clients across 3 product lines (SFA, eB2B, COE), translating business requirements into client-specific KPI logic.

Won the Lead Performer Award (2025) for independently designing and delivering the platform's custom KPI computation logic.

JavaNettyElasticsearchRedisKafkaAWS S3MySQLJenkinsREST APIsCI/CD

CONTINUOUS_INTEGRATION: IN_PROGRESS

SYSTEMSENGINEERING

03_REPOSITORIES.SYS

Status: Production-Grade Systems Active

COMPLETED

Socratese

// A Socratic tutor over Obsidian notes: a local RAG pipeline grounds an LLM that only asks questions, never answers, forcing you to reconstruct your own notes from memory.

Technical Challenge

Resolved 6-turn question loops by moving stall detection into deterministic code after 3 measured prompt fixes failed (1 made adherence worse); shipped 249 mutation-verified tests with the TUI driven headlessly under strict pyright.

PythonTyperTextualRAGOpenAI EmbeddingsChromaAnthropic ClaudeSQLite

COMPLETED

InsightsHub

// An AI-powered GitHub analytics engine transforming activity into narrative insights using LLMs and async data pipelines.

Technical Challenge

Managed high-volume API rate limiting and data consistency by implementing a Redis-backed asynchronous queue and GraphQL batching.

PythonFastAPIPostgreSQLRedisLLM IntegrationGraphQLNext.js

COMPLETED

Event Processing Platform

// A production-grade event-driven system implementing idempotent consumers and comprehensive observability pipelines with Prometheus and Grafana.

Technical Challenge

Ensured exactly-once processing using Redis-based deduplication and built a resilient architecture with Dead Letter Queue (DLQ) handling and real-time Kafka consumer lag monitoring.

JavaSpring BootKafkaPostgreSQLRedisPrometheusGrafana

Labs & Learning

COMPLETED

Music Library API

// A RESTful API for managing music libraries with JWT authentication and fuzzy search.

GoGinPostgreSQLJWTDocker

COMPLETED

Personal Shell

// A custom command-line interface exploring low-level system programming and process management.

GoCLISystem Programming

CONTINUOUS_DEPLOYMENT: OPERATIONAL

Connect

Connect.exe

SYSTEM STATUS: ACTIVE | --:--:--

contact.yml

# contact.yml

service: tushar-sharma

status: open_to_work

remote: true

runtime:

role: backend engineer

stack: [python, fastapi, java, netty]

data: [postgres, redis, kafka, elasticsearch]

healthcheck:

path: /ping

interval: 30s

$ curl -s localhost:8080/ping

{"status": "ok", "open_to_work": true}_

# Note: Always looking for challenges that require deep technical knowledge and creative problem solving.

TusharSharma

01_PROFILE.MD

DECODING COMPLEXITY. SCALING SOLUTIONS.

Core_Stack

Process

04_SYSTEM_DEPENDENCIES

02_EXPERIENCE.EXE

Software Engineer

03_REPOSITORIES.SYS

Socratese

InsightsHub

Event Processing Platform

Music Library API

Personal Shell

Connect.exe

DECODING COMPLEXITY.
SCALING SOLUTIONS.