Najath Akram
Signal Processing Engineer, Airspan Networks / Houston, Texas
I build the part of a radio you never see.
These days that means multiband O-RAN radio units at Airspan, some early 6G, and AI agents that have to prove their work like everyone else.
The physical layer: where information becomes electromagnetic waves and, with a bit of luck, becomes information again on the other end.
Who is writing
Hello. I'm Najath. Ten years in radio, the last six shipping O-RAN radio units: a PhD at Florida International University, then Jabil, where I went from Senior FPGA Engineer to Lead to Principal in four years, and now Airspan.
The work lives somewhere between math and hardware. Most of my days are spent in MATLAB, in front of a spectrum analyzer, or in a document describing why a particular combination of frequencies decided to interfere with itself.
The models are half the job. The other half is the seam between the radio and the people shipping it: since February 2025 I have taken 27 production debug investigations down to root cause on lab and field captures, and packaged the analyzers as standalone executables so the lab teams do not need me in the room.
Some of that I now build with AI agents in the loop, which changed how much I write and not one thing about what I trust. The verification came first and the agents arrived later: 846 automated checks on the PRACH chain, independent decoders that re-read every exported byte off disk, mutation tests that plant faults to prove the checks can see them. That seam is the part I keep choosing, and the rest of this site is the case for it.
What I work on
Signal processing
Digital beamforming, phased arrays, and the quiet art of doing more with fewer ADCs. A good portion of my PhD went into trading converters for mathematics: one ADC per four antennas at 28 GHz, and half the count again for 2D arrays, using multidimensional signal processing and approximate FFTs.
O-RAN radio units
Digital front ends, crest factor reduction, EVM, PIM, and the fronthaul that carries it all: split 7-2x, eAxC, the WG4 CUS-plane. I spend most of my days here. If something about a waveform looks wrong on a spectrum analyzer, it usually becomes my afternoon.
Machine learning
Machine learning for wireless, applied carefully: a polyphase filter bank feeding a signal classifier on NVIDIA Orin, and a current experiment in letting AI agents near radio-unit verification, where a judge side and an implementer side are kept apart on purpose and no completion claim counts until a human countersigns it.
Recent work
- Modeling
The golden models a radio is signed off against
Bit-accurate models of the downlink, uplink, and PRACH chains of multiband O-RAN radio units. Five vendor fixed-point IP C-models wrapped through MEX, so the model runs the same arithmetic the FPGA does, then exports every pipeline stage for verification engineers to diff against RTL and fronthaul captures. Seventeen bands, carriers from 3 to 100 MHz, and versioned executables so nobody needs a MATLAB license to run them.
- Verification
A PRACH chain that checks itself 846 times
PRACH is where a radio first hears a phone, so it is worth being paranoid about. The model covers LTE formats 0 to 3 and full NR FR1, long and short preambles, all 256 configuration indices in both FDD and TDD, proven by an exhaustive 512-case sweep. A harness re-decodes every exported byte back off disk through independently written decoders, including a blind Zadoff-Chu root search. 846 checks across 31 configurations; 29 clean, and 2 that still fail on a documented window-placement defect I have not fixed yet. Mutation testing caught all 10 faults I deliberately planted.
- Research
Uplink combining for shared cells
When several ceiling-mounted radios hear the same phone, someone has to combine them well. A fixed-point model of the firmware datapath, up to four radios by eight antennas, matching the floating-point reference within 0.01 dB SINR across a twelve-configuration matrix. On the real vendor FFT front end, maximum-ratio combining holds 34 dB SINR in the demo where equal gain collapses below 8. The useful discovery was algebraic: combining inside each radio first and then across them loses exactly 0.000 dB, so each radio forwards one stream instead of eight. That is the whole bandwidth argument for the architecture.
- Tooling
A waveform analyzer for the lab bench
A desktop app that decodes O-RAN fronthaul captures into uplink resource grids and measures PUSCH EVM for NR and LTE against the same configuration files the reference instrument uses. It also finds and demodulates format-0 PRACH preambles straight out of a capture, validating the recovered root and cyclic shift against the test setup. Bench-verified around 1.8% EVM at 20 MHz and 7.1% at 100 MHz, and it ships as one standalone executable.
Papers
IEEE Access, vol. 9
IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 68
IEEE Access, vol. 8
If you are hiring for product or technical program roles in wireless, or you have a good question about O-RAN fronthaul, uplink combining, or where machine learning belongs in a radio, email reaches me directly. I read everything.
Product and technical program roles in wireless, where radio depth and a shipped-product record both count.