Abhinav Kumar profile photo 1 on Tal

Abhinav Kumar

Full Stack Software Engineer | Building production Python APIs & ML pipelines at Skylark Drones | TypeScript | Django | DevOps | AWS | Microservices

Bengaluru

5+ years of work experience

Skylark Drones
Jul 20242 yrs 3 mosPresent

Bengaluru

• Redesigned a folder-linking UI refresh strategy for 18,000+ drone images to be subtree-aware — only refetching when the affected folder intersects the current view. Load time: 45s → 8s. Not a patch — an architectural decision with the help of AI driven development workflows • Designed and built 4+ Python microservices from scratch (FastAPI/Flask) with fault isolation, async patterns, and end-to-end AWS integration — each service independently deployable so a failure in one doesn't cascade to others. • Eliminated N+1 query bottlenecks across core Django REST APIs using select_related and prefetch_related — reducing database round-trips from 37 to 5 and significantly improving production PostgreSQL response latency. • Migrated ML inference workloads from CPU to GPU-backed EC2 instances; designed benchmarking utilities tracking spin-up time, RAM utilisation, spot interruption rates, and cost per inference — resulting in 25% infrastructure cost reduction at equivalent model performance. • Implemented an AWS spot instance capacity fallback mechanism (InsufficientInstanceCapacity handling) — eliminating production ML pipeline outages caused by spot unavailability. • Acted as cross-product integration lead between two internal platforms: designed secure REST API contracts, implemented API-key authorisation, and provided end-to-end onboarding to the partner engineering team. • Served as Release Coordinator across 5 production cycles — owning deployment scheduling, stakeholder communication, and real-time issue triage with zero-delay rollouts. • Reduced CI/CD deployment time by 50% by replacing remote GitLab shared runners with locally configured dedicated runners, eliminating queue wait and cold-start cache misses. • Led on-premises deployment of a critical microservice with Nginx reverse proxy and SSL (Certbot) — extending the platform's enterprise readiness to air-gapped client environments.
Techshots
Nov 20232 yrs 11 mosPresent

Bengaluru

Present Building a tech education channel on Instagram with 40,000+ followers — focused on making software engineering concepts, DSA, and system design genuinely simple to understand. Started it as a way to force myself to understand things deeply enough to explain them clearly. That discipline — translate the complex into the simple — shapes how I write documentation, give demos, and communicate technical decisions to non-technical stakeholders. Format: short-form video (Reels). Niche: developers at all levels.
Skylark Drones
Jan 20247 mosJul 2024

Bangalore Urban

Integrated Meta's Segment-Anything Model (SAM) into production drone imagery annotation workflows — increasing throughput by 90% by replacing manual polygon drawing with AI-assisted segmentation. Handled latency via async inference queuing and added confidence-threshold review gates to prevent low-quality predictions from entering training data. • Built a Mapbox-powered browser-based annotation interface for custom ML models — reducing annotation cycle times from days (desktop GIS tools + file handoffs) to minutes via direct streaming from the platform. • Designed dynamic Vue.js components in TypeScript for real-time telemetry updates using WebSocket connections — type-safe payloads ensured schema changes in drone data didn't silently break the UI.
Easebuzz
Jun 20233 mosAug 2023

Pune District

• Improved deployment efficiency by 40% through CI/CD pipeline optimisations with Jenkins — parallel stage execution, dependency caching, and artifact reuse across builds. • Architected ELK stack + Grafana integration for centralised log aggregation and real-time infrastructure monitoring across microservices deployed on AWS (EC2, Lambda, RDS).
Parikshit Student Satellite
Dec 20229 mosAug 2023

Udipi, Karnataka, India

Parikshit Student Satellite
Jan 20212 yrsDec 2022

Udipi, Karnataka, India

Parikshit Student Satellite
Nov 20203 mosJan 2021

Udipi, Karnataka, India

Mind Commute
Sep 20203 yrsAug 2023

- Built and developed an innovative mental health website, orchestrating a captivating front-end interface through the integration of HTML, CSS5, JS, and Bootstrap. - Orchestrated a dynamic user experience powered by the cutting-edge Python Django Framework, catalyzing an astonishing 100+ registered users in user engagement and invaluable feedback.

Abhinav’s AI Stack

antigravity
antigravity
claude code
claude code
claude
claude
codex
codex
gemini
gemini
github copilot
github copilot

Education

Manipal institute of technology

Bachelor of Technology - BTech, EIE

Manipal institute of technology

20204 yrs2024

High School Diploma, PCM with CS

The Khaitan School

20164 yrs2020

Join Abhinav and 10k+ other professionals on

Tal Boss

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

Download Tal Boss

Abhinav Kumar on Tal

Headline

Full Stack Software Engineer | Building production Python APIs & ML pipelines at Skylark Drones | TypeScript | Django | DevOps | AWS | Microservices

About

I build systems that own their outcomes — not just their uptime. At Skylark Drones, I've spent the last 2+ years as a full-stack engineer with end-to-end ownership of a production drone intelligence platform: Python microservices on AWS Fargate, a Vue frontend handling 18,000+ image datasets, ML inference pipelines on GPU-backed EC2, and five production release cycles as coordinator. Some things I've shipped that I'm proud of: → Redesigned a folder-linking refresh architecture to be subtree-aware — 45s load time down to 8s without changing the backend API contract → Eliminated N+1 query bottlenecks across core REST APIs — 37 database round-trips down to 5 via Django ORM optimisation → Integrated Segment-Anything (SAM) into production annotation workflows — 90% throughput increase, async inference with confidence-threshold review gates → Built benchmarking infrastructure across EC2 instance families — 25% infra cost reduction at same model performance Beyond the code: I've led cross-product integrations between separate engineering teams, given demos to enterprise clients, mentored interns, and coordinated zero-delay production rollouts under real pressure. I also run TechShots60 — a tech education channel with 40,000+ followers — because the fastest way to understand something deeply is to explain it simply. Based in Bengaluru. Contact: abhinav.kumar2712@gmail.com

Location

Bengaluru

AI stack

antigravity

claude code

claude

codex

gemini

github copilot

Work experience

Software Engineer at Skylark Drones, Jul 2024 - Present · 2 yrs 3 mos, Bengaluru, • Redesigned a folder-linking UI refresh strategy for 18,000+ drone images to be subtree-aware — only refetching when the affected folder intersects the current view. Load time: 45s → 8s. Not a patch — an architectural decision with the help of AI driven development workflows • Designed and built 4+ Python microservices from scratch (FastAPI/Flask) with fault isolation, async patterns, and end-to-end AWS integration — each service independently deployable so a failure in one doesn't cascade to others. • Eliminated N+1 query bottlenecks across core Django REST APIs using select_related and prefetch_related — reducing database round-trips from 37 to 5 and significantly improving production PostgreSQL response latency. • Migrated ML inference workloads from CPU to GPU-backed EC2 instances; designed benchmarking utilities tracking spin-up time, RAM utilisation, spot interruption rates, and cost per inference — resulting in 25% infrastructure cost reduction at equivalent model performance. • Implemented an AWS spot instance capacity fallback mechanism (InsufficientInstanceCapacity handling) — eliminating production ML pipeline outages caused by spot unavailability. • Acted as cross-product integration lead between two internal platforms: designed secure REST API contracts, implemented API-key authorisation, and provided end-to-end onboarding to the partner engineering team. • Served as Release Coordinator across 5 production cycles — owning deployment scheduling, stakeholder communication, and real-time issue triage with zero-delay rollouts. • Reduced CI/CD deployment time by 50% by replacing remote GitLab shared runners with locally configured dedicated runners, eliminating queue wait and cold-start cache misses. • Led on-premises deployment of a critical microservice with Nginx reverse proxy and SSL (Certbot) — extending the platform's enterprise readiness to air-gapped client environments.

Content Creator at Techshots, Nov 2023 - Present · 2 yrs 11 mos, Bengaluru, Present Building a tech education channel on Instagram with 40,000+ followers — focused on making software engineering concepts, DSA, and system design genuinely simple to understand. Started it as a way to force myself to understand things deeply enough to explain them clearly. That discipline — translate the complex into the simple — shapes how I write documentation, give demos, and communicate technical decisions to non-technical stakeholders. Format: short-form video (Reels). Niche: developers at all levels.

SDE intern at Skylark Drones, Jan 2024 - Jul 2024 · 7 mos, Bangalore Urban, Integrated Meta's Segment-Anything Model (SAM) into production drone imagery annotation workflows — increasing throughput by 90% by replacing manual polygon drawing with AI-assisted segmentation. Handled latency via async inference queuing and added confidence-threshold review gates to prevent low-quality predictions from entering training data. • Built a Mapbox-powered browser-based annotation interface for custom ML models — reducing annotation cycle times from days (desktop GIS tools + file handoffs) to minutes via direct streaming from the platform. • Designed dynamic Vue.js components in TypeScript for real-time telemetry updates using WebSocket connections — type-safe payloads ensured schema changes in drone data didn't silently break the UI.

Devops Intern at Easebuzz, Jun 2023 - Aug 2023 · 3 mos, Pune District, • Improved deployment efficiency by 40% through CI/CD pipeline optimisations with Jenkins — parallel stage execution, dependency caching, and artifact reuse across builds. • Architected ELK stack + Grafana integration for centralised log aggregation and real-time infrastructure monitoring across microservices deployed on AWS (EC2, Lambda, RDS).

Senior Communication Subsystem Member at Parikshit Student Satellite, Dec 2022 - Aug 2023 · 9 mos, Udipi, Karnataka, India

Team Member at Parikshit Student Satellite, Jan 2021 - Dec 2022 · 2 yrs, Udipi, Karnataka, India

Taskphase Recruit at Parikshit Student Satellite, Nov 2020 - Jan 2021 · 3 mos, Udipi, Karnataka, India

Full-stack Developer at Mind Commute, Sep 2020 - Aug 2023 · 3 yrs, - Built and developed an innovative mental health website, orchestrating a captivating front-end interface through the integration of HTML, CSS5, JS, and Bootstrap. - Orchestrated a dynamic user experience powered by the cutting-edge Python Django Framework, catalyzing an astonishing 100+ registered users in user engagement and invaluable feedback.

Education

Bachelor of Technology - BTech, EIE, Manipal institute of technology

High School Diploma, PCM with CS, The Khaitan School

Public links

github: https://github.com/abhinav2712

Abhinav Kumar - Software Engineer in Bengaluru | Tal