Harsh Tiwari profile photo 3 on Tal
Harsh Tiwari profile photo 1 on Tal
Harsh Tiwari profile photo 2 on Tal
Harsh Tiwari profile photo 3 on Tal
Harsh Tiwari profile photo 1 on Tal

Harsh Tiwari

Software Engineer | Bachelor of Technology in Computer Science

Bengaluru

5+ years of work experience

WellnessLiving
Apr 20251 yr 3 mosPresent

Bengaluru, Karnataka, India

Engineered a multi-tenant Voice AI backend with a generalized function-calling layer to orchestrate LLM-driven actions across provider integrations, enabling extensible tenant-specific workflows through RAG and tool-execution pipelines for production voice agents. Re-architected Cerebrum, our LLM microservice, to deliver sub-second telephony responses by streaming model output, parallelizing retrieval and tool preparation, trimming prompt-construction overhead, and removing synchronous bottlenecks from the inference path. Scaled a WebSocket platform with stateless services, OAuth 2.0 authentication, and rate limiting, improving connection stability and reducing customer-reported issues in real-time chat across tenant-facing workflows. Operationalized Airbyte ingestion pipelines from systems including Intercom and Zoom into Redshift, improving the reliability of centralized analytics and downstream reporting. Launched a RAG-backed analytics interface with SQL agents to enable self-serve reporting for internal teams and reduce dependence on manual data requests for recurring business queries. Deployed production-grade latency instrumentation across 5 stages of the voice path: STT, request assembly, LLM inference, tool execution, and TTS, using structured tracing and service-level timing hooks, streaming telemetry through Kafka into Redshift for tail-latency attribution and sub-second response optimizations.
kusho
Aug 20231 yr 9 mosApr 2025

Built Redis and RQ-based worker pipelines for scalable LLM task execution, prompt processing, and asynchronous job orchestration across large API payloads, improving throughput under heavy background workloads. Designed a unified service-definition DSL for 4 API specification formats: Postman, OpenAPI/Swagger, GraphQL, and Protocol Buffers, then integrated LLM APIs to auto-generate test cases and reduce test authoring time. Built an API testing and execution platform with 50+ workflow node types for end-to-end orchestration, improving large-flow execution performance and shortening test runtime for complex workflows. Developed a Chrome extension that records user interactions and generates Playwright test scripts, implemented robust selectors with fallback strategies for UI changes, and delivered 95% stability, 3x test coverage, and 70% less manual QA. Built a distributed test execution engine that modeled large API collections as dependency-aware workflow graphs, executing nodes via Redis based worker queues (RQ) with dependency-aware scheduling, parallel fan-out, and isolated execution pipelines to support high-concurrency workflow simulation across 1000+ businesses. Built a compiler that normalized 4 specification formats including Postman, OpenAPI, GraphQL, and Protobuf into a common executable graph IR, enabling a shared runtime for test generation, validation, and large-scale API workflow execution.
Castler
May 20212 yrs 4 mosAug 2023

India

• Developed high-throughput payment system with Redis and circuit breakers for UPI/IMPS/NEFT, processing 100Cr+ monthly with 99.99% uptime. • Implemented bank-grade security using OAuth 2.0/PKCE with automated scanning for PCI DSS compliance. • Led zero-downtime migration of 500GB+ financial data from MongoDB to MySQL, achieving 50% faster queries and RBI compliance.
Wikilimo
Dec 20202 mosJan 2021

* Developed machine learning models for weather forecasting. * Improved accuracy of models by 79% by using CNNs together with LSTMs. * Generated highly detailed weather anomaly reports, providing critical insights.

Harsh’s AI Stack

claude code
claude code
cursor
cursor
deepseek
deepseek
elevenlabs
elevenlabs
hermes agent
hermes agent
hugging face
hugging face

Education

amity University

Bachelor of Technology, Computer Science

amity University

20174 yrs2021

Join Harsh and 10k+ other professionals on

Tal Boss

Tal is the app where bosses meet engineers & builders directly, without middlemen

Download Tal Boss

Harsh Tiwari on Tal

Headline

Software Engineer | Bachelor of Technology in Computer Science

About

🚀 Software Engineer with 3+ years of experience specializing in full-stack development and machine learning engineering. Passionate about building scalable solutions and optimizing system performance. 💻 Technical Expertise: • Full-stack development with React.js, Node.js, and MongoDB • Machine learning model development and optimization • Performance tuning and database optimization • Microservices architecture and API integration 🎯 Key Achievements: • Optimized MongoDB queries leading to 70% reduction in response time • Achieved 99.9% uptime for critical banking integrations • Improved ML model accuracy by 79% using CNN-LSTM hybrid architecture • Led frontend development for enterprise platforms 🔍 Currently focusing on: • Building innovative solutions at KushoAI • Exploring cutting-edge technologies in ML/AI • Contributing to open-source projects 💡 I believe in continuous learning and sharing knowledge with the tech community. Active contributor to technical discussions and mentor to aspiring developers.

Location

Bengaluru

AI stack

claude code

cursor

deepseek

elevenlabs

hermes agent

hugging face

Work experience

Senior Software Engineer at WellnessLiving, Apr 2025 - Present · 1 yr 3 mos, Bengaluru, Karnataka, India, Engineered a multi-tenant Voice AI backend with a generalized function-calling layer to orchestrate LLM-driven actions across provider integrations, enabling extensible tenant-specific workflows through RAG and tool-execution pipelines for production voice agents. Re-architected Cerebrum, our LLM microservice, to deliver sub-second telephony responses by streaming model output, parallelizing retrieval and tool preparation, trimming prompt-construction overhead, and removing synchronous bottlenecks from the inference path. Scaled a WebSocket platform with stateless services, OAuth 2.0 authentication, and rate limiting, improving connection stability and reducing customer-reported issues in real-time chat across tenant-facing workflows. Operationalized Airbyte ingestion pipelines from systems including Intercom and Zoom into Redshift, improving the reliability of centralized analytics and downstream reporting. Launched a RAG-backed analytics interface with SQL agents to enable self-serve reporting for internal teams and reduce dependence on manual data requests for recurring business queries. Deployed production-grade latency instrumentation across 5 stages of the voice path: STT, request assembly, LLM inference, tool execution, and TTS, using structured tracing and service-level timing hooks, streaming telemetry through Kafka into Redshift for tail-latency attribution and sub-second response optimizations.

Full Stack Engineer at kusho, Aug 2023 - Apr 2025 · 1 yr 9 mos, Built Redis and RQ-based worker pipelines for scalable LLM task execution, prompt processing, and asynchronous job orchestration across large API payloads, improving throughput under heavy background workloads. Designed a unified service-definition DSL for 4 API specification formats: Postman, OpenAPI/Swagger, GraphQL, and Protocol Buffers, then integrated LLM APIs to auto-generate test cases and reduce test authoring time. Built an API testing and execution platform with 50+ workflow node types for end-to-end orchestration, improving large-flow execution performance and shortening test runtime for complex workflows. Developed a Chrome extension that records user interactions and generates Playwright test scripts, implemented robust selectors with fallback strategies for UI changes, and delivered 95% stability, 3x test coverage, and 70% less manual QA. Built a distributed test execution engine that modeled large API collections as dependency-aware workflow graphs, executing nodes via Redis based worker queues (RQ) with dependency-aware scheduling, parallel fan-out, and isolated execution pipelines to support high-concurrency workflow simulation across 1000+ businesses. Built a compiler that normalized 4 specification formats including Postman, OpenAPI, GraphQL, and Protobuf into a common executable graph IR, enabling a shared runtime for test generation, validation, and large-scale API workflow execution.

Full Stack Engineer at Castler, May 2021 - Aug 2023 · 2 yrs 4 mos, India, • Developed high-throughput payment system with Redis and circuit breakers for UPI/IMPS/NEFT, processing 100Cr+ monthly with 99.99% uptime. • Implemented bank-grade security using OAuth 2.0/PKCE with automated scanning for PCI DSS compliance. • Led zero-downtime migration of 500GB+ financial data from MongoDB to MySQL, achieving 50% faster queries and RBI compliance.

Machine Learning Engineering Intern at Wikilimo, Dec 2020 - Jan 2021 · 2 mos, * Developed machine learning models for weather forecasting. * Improved accuracy of models by 79% by using CNNs together with LSTMs. * Generated highly detailed weather anomaly reports, providing critical insights.

Education

Bachelor of Technology, Computer Science, amity University

Public links

tiwariaiharsh.com: https://www.tiwariaiharsh.com/